Skip to content

Latest commit

 

History

History
409 lines (245 loc) · 13.1 KB

CHANGELOG.md

File metadata and controls

409 lines (245 loc) · 13.1 KB
Raw

Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

Unreleased

0.36.0 - 2022-09-12

Added

  • input multiplicity within the cli in debug mode
  • list_model to the SDK client

Removed

  • chore: remove the examples section, and references to them

Changed

  • BREAKING CHANGE: opener input identifier must be opener
  • BREAKING CHANGE: datasample input identifier must be datasamples
  • BREAKING CHANGE: chainkeys, opener and datasamples are forbidden identifier for inputs and outputs of kind artifact and performance
  • BREAKING CHANGE: pass the algo method to execute under the --method-name argument within the CLI. If the interface between substra and the backend is handled via substratools, there are no changes to apply within the the substra code but algo and metric Dockerfiles should expose a --method-name argument in the ENTRYPOINT (#266)
  • test tasks now have an higher rank than their parent train task
  • chore: remove close source for CHANGELOG, README and cyclic-example
  • chore: rename connect-tools to substra-tools.
  • BREAKING CHANGE: remove the shared local folder of the compute plan
  • feat: download function of the client return the path of downloaded file.

Feature

  • Local mode: add a check, a task output of type performance must have public permissions

Fixes

  • Filters: Fix the filters on status for compute plans and tasks

This fix also introduces some changes: the value for the filters on status must now be a list (like for other filters, there is a OR condition between elements of the list) and its value must be substra.models.ComputePlanStatus.{name of the status}.value for compute plans and substra.models.Status.{name of the status}.value for tasks.

Example:

# Return all the compute plans that are not finished
client.list_compute_plan(filters={'status':[substra.models.ComputePlanStatus.todo.value, substra.models.ComputePlanStatus.waiting.value, substra.models.ComputePlanStatus.doing.value]})

# Return all the composite traintuples with the status "doing"
client.list_composite_traintuple(filters={"status": [substra.models.Status.doing.value]})

Documentation

  • Add the filters documentation to the references

0.35.0 - 2022-08-29

Changed

  • chore: change the metrics and algo definition relying on substra tools. All the methods of those objects now takes inputs and outputs as arguments; which are TypedDict.

0.34.0 - 2022-08-22

Changed

  • BREAKING CHANGE:in local mode, each client has its own organization_id. Removed the DEBUG_OWNER mechanism.

Instead of:

client = substra.Client(debug=True)
clients = [client] * 2

do:

clients = [substra.Client(debug=True) for _ in range(2)]
client1_org_id = clients[0].organization_info().organization_id
  • The Client.organization_info function now returns a model OrganizationInfo instead of a dict

0.33.0 - 2022-08-17

Added

  • New update_compute_plan, update_algo and update_dataset methods that allow editing names
  • The local mode now uses the inputs/outputs fields to execute the tasks. The other fields are still kept up-to-date
  • Support for the transient flag on tasks outputs

Changed

  • BREAKING CHANGE: rename schemas.ComputeTaskOutput to schemas.ComputeTaskOutputSpec

Removed

  • BREAKING CHANGE: remove the chainkey feature

0.32.0 - 2022-08-09

Added

  • BREAKING CHANGE: drop Python 3.7 support
  • add inputs and outputs fields to the Algo model

Removed

  • BREAKING CHANGE CLI: keep only the cancel, profile, login and organization commands

Changed

  • BREAKING CHANGE: Stop using legacy _permissions fields

0.31.0 - 2022-08-01

Added

  • "inputs" field to substra.sdk.schemas.*tupleSpec
  • "inputs" field to *tuple models
  • embed models/performances in output fields of tasks

0.30.1 - 2022-07-26

Fixed

  • revert: fix: get_performances works again in remote

0.30.0 - 2022-07-25

Fixed

  • fix: get_performances works again in remote

Changed

  • feat: change DEFAULT_BATCH_SIZE to 500

Added

  • "outputs" field to substra.sdk.schemas.*tupleSpec
  • feat: add "outputs" field to *tuple models

Removed

  • BREAKING: "out_trunk_model_permissions" field from substra.sdk.schemas.CompositeTraintupleSpec (superseded by "outputs")

0.29.0 - 2022-07-11

Added

  • feat: support two composite as input of a composite in local mode
  • feat: support metadata filters in local mode

Changed

  • BREAKING: convert (test task) to (predict task + test task)

Fixed

  • fix: add mandatory cp name in cyclic strategy example

0.28.0 - 2022-07-05

Changed

  • BREAKING CHANGE: Client.update_compute_plan renamed Client.add_compute_plan_tuples

Removed

  • CLI commands add, get and list
  • BREAKING: Removed metrics APIs and types; use algo APIs and types instead

0.27.0 - 2022-06-27

Changed

  • BREAKING: implement filtering and ordering for list methods in SDK

    • Removed filtering syntax asset::key::value in SDK and CLI

    • New filtering syntax in SDK filters={key:["value1", "value2"]} and ordering possibility:

      - list_***(self, filters=None) -> List[substra.sdk.models.***]
+ list_***(self, filters: dict = None, order_by: str = 'creation_date', ascending: bool = False) -> List[substra.sdk.models.***]

0.26.0 - 2022-06-20

Fixed

  • fix: get_performance test tuples filter on local dal

0.25.0 - 2022-06-14

Fixed

  • cancel_compute_plan should return nothing

    SDK:

    - cancel_compute_plan(self, key: str) -> substra.sdk.models.ComputePlan
+ cancel_compute_plan(self, key: str) -> None

- Cancel execution of compute plan, the returned object is described in the [models.ComputePlan](sdk_models.md#ComputePlan) model
+ Cancel execution of compute plan. Nothing is returned by this method

    CLI:

    substra cancel compute_plan [OPTIONS] COMPUTE_PLAN_KEY
-  Cancel execution of a compute plan.
+  Cancel execution of a compute plan. Nothing is printed, you can check again the compute plan status with `substra get compute_plan`.

Changes

  • BREAKING CHANGE (rename node to organization)
    • SDK: client.list_node renamed to client.list_organization
    • SDK: client.node_info renamed to client.organization_info
    • CLI: the command substra node info is now substra organization info
    • CLI: the command substra list node is now substra list organization
    • local mode: when using chainkeys, the chainkey file is renamed from node_name_id.json to organization_name_id.json

0.24.0 - 2022-06-07

Added

  • feat: add predict algo category

Changes

  • BREAKING CHANGE (UpdateComputePlanSpec key field): Use new register tasks endpoint

0.23.0 - 2022-05-31

Added

  • feat: add empty compute plan status

Changed

  • feat: in debug mode, subprocess errors are now caught and raised as ExecutionError

0.22.0 - 2022-05-22

Added

  • feat: add round_idx to get_performance()

Changed

  • feat: Add algo named inputs and outputs for existing task in remote mode
  • feat: log level set to warning for substra-tools

0.21.0 - 2022-05-16

Changed

  • BREAKING CHANGE: add mandatory name field to compute plan
  • doc: update orchestrator compatibility table for the Substra release 0.11.0

Added

  • feat: save performances in live for local mode on a json file

0.20.0 - 2022-05-09

Changed

  • BREAKING CHANGE: use algo spec for metric
  • BREAKING CHANGE: Cancel CP returns nothing
  • BREAKING CHANGE: the user must give the key of the compute plan when creating a new compute plan

0.19.0 - 2022-05-03

Added

  • feat(sdk): add get_datasample method
  • feat: add the get_performances(compute_plan.key) function to the client

Changed

  • BREAKING CHANGE: local mode - arguments of each train, test, composite and aggregate task are passed as named inputs and outputs in a json format. Those changes are compatible with substra-tools 0.12.0.
  • Rest client supports default backend pagination, list will automatically fetch all pages
  • BREAKING CHANGE: local mode - the format of the asset ids is now uuid.UUID, e.g. 3da0075f-360b-446d-902f-af60a38d9594 and not the hex representation.
  • update substra-tools version in examples Dockerfiles

0.18.0 - 2022-04-11

Changed

  • Local mode - in a compute plan, the testtuples are executed at the same time as the other tasks instead of after all of them
  • Examples now use the substra-tools base Docker image 0.10.0

Fixed

  • Local mode works with data on another disk partition than the one the code is executed on
    • In subprocess mode: the data samples are linked via a symbolic link to a temporary folder
    • In Docker mode: the data samples are copied at the task execution time to a temporary folder
  • Be able to use pdb in subprocess mode

0.17.0 - 2022-03-01

Changed

  • Speed and disk space usage improvement as datasamples are not copied to local backend folders anymore
  • Documentation: Fix the display of the API reference

Added

  • CLI: Add the logs command to display and download the logs of a failed tuple
  • SDK: Add the get_logs and download_logs methods
  • doc: add DEBUG_SPAWNER doc
  • doc: add AlgoSpec to documentation
  • SDK: add get_model method

Fixed

  • Local mode should not accept test data samples for train tuples

0.16.0 - 2022-01-14

Changed

  • (BREAKING) Add a logs_permission field to Dataset schema and model

Fixed

  • CLI: Fix the display of metric key and perfs for 'substra get testtuple'

Removed

  • (BREAKING) Remove the download permissions

0.15.0 - 2022-01-10

Changed

  • (BREAKING) Remove the datasample.validated field as it's deprecated

Fixed

  • CLI: fix the display of the assets in yaml and json format
  • No more crash when running a script in local subprocess mode from a path with spaces
  • Examples: add dataset logs_permission field

Added

  • Error type to traintuple, testtuple, aggregatetuple and composite_traintuple
  • Expose the backend type

0.14.0 - 2021-12-01

Added

  • Cyclic strategy example

Fixed

  • Local mode - the substra specific assets are now saved in /substra_internal instead of the work dir
  • Fix the display of the composite traintuple if it has no input models
  • Properly prevent path traversal in archives and don't allow symbolic links

0.13.0 - 2021-11-02

Added

  • traintuple can take aggregatetuple as in_tuple
  • Pretty print for Model and Spec
  • Display orchestrator and chaincode versions in node info command
  • Accept full datamanager, metrics and parent_tasks objects in get_*_tuple responses
  • Add task extra information related to start_date and end_date

Changed

  • Ignore extra fields in API response
  • (BREAKING) Replace objective by metric
  • (BREAKING) Multiple metrics and performances per test task

Improvements

  • Local mode, execution in Docker: re-use the Docker images so that the execution is faster

0.11.0 - 2021-10-04

Added

  • Display backend version in node info command

Changed

  • Internal: backend now serves paginated lists of asset. Client still returns simple lists as before.
  • (BREAKING CHANGE) All algos are now one single object with a category property.
  • (BREAKING CHANGE) Various API changes due to the architecture change introduced by the orchestrator.

0.10.0 - 2021-08-05