Releases: materialsproject/jobflow
Releases · materialsproject/jobflow
v0.1.9
New features
- Delayed updates to config and metadata for dynamic flows. See docstring of
Job.update_metadata
for more details (@gpetretto, #198) - Additional stores are now generated on the fly as memory stores if they are not
specified in jobflow settings (@davidwaroquiers, #183)
Bug fixes:
- Optimised calls to update_kwargs (@jmmshn, #177)
- "job_uuid" and "job_index" are now indexed fields in additional stores (@jmmshn, #165)
- Fix additional store storing
None
(@mjwen, #160)
Enhancements:
v0.1.8
New features:
- New
update_metadata
function for updating the metadata of jobs and flows. - New
update_config
function for updating the config (included manager_config) of
jobs and flows. - New
DIRECTORY_FORMAT
option in JobflowSettings for controlling the date time format
used to create new directories. - New functions for adding and removing jobs from a flow. The
Job.jobs
list is no
longer mutable (@gpetretto). - New
Job.hosts
attribute that stores a list of all host Flows. This captures the
nested nature of flows with the outer flow always first in the list (@gpetretto).
Bug fixes:
v0.1.7
New features:
- Validate subschemas of nested models (@gpetretto, #118).
downstream_manager_config
for controlling config of dynamic jobs (@arosen93, #121).- S3Store yaml parsing (@jmmshn, #124).
Bug fixes:
- Fix home path for loading settings (@gpetretto, #119).
- Docs updates (@arosen93, #111).
v0.1.6
v0.1.5
Bug fixes:
- Remove
JobConfig.pass_metadata
option and instead pass metadata automatically. - Fix serialization compatability with the FireWorks workflow manager.
v0.1.4
New features:
- Add
append_name
option toJob
andFlow
that allow easy modification of all
job names in a flow. - Add
JobConfig.pass_metadata
(defaults to True) that can be used to pass job metadata
on to dynamically added jobs.
Bug fixes:
- Fireworks manager now adds metadata to FireWork spec. Fixes #21.
v0.1.3
Jobflow now uses pydantic to handle settings. Currently, there is only a single setting
JOB_STORE
which controls the default store used by run_locally
and the fireworks
manager. You can update the default store by writing a ~/.jobflow.yaml
settings
file. See the API documentation for more details.
v0.1.2
New features:
ensure_success
option added torun_locally
.- Better graph visualisation.
- Updating the name of a job from a maker now propogates the name change to the maker.
Job.update_maker_kwargs
withnested=True
now applies the updates to makers
in the kwargs or args of the job.
v0.1.1
Docs updates.
v0.1.0
Major changes:
Schema
class removed. Any pydantic model can now be an output schema.
Enhancements:
JobStore.get_output
now resolves references in the output of other jobs.JobStore.get_output
:which
now supports specifying a specific job index.- Better support for circular and missing references in
JobStore.get_output
and
OutputReference.resolve
. - Update dependencies to use latest jsanitize features.
Bug fixes:
- Fixed issue with references in flow of flows (@davidwaroquiers, #18).
- Makers now allow non-default parameters (fixes: #13).
- Fix reference cache with multiple indexes.