- BF: Update docker setup and dependencies (#307)
- DOC: Update minimal required permissions (#302)
- BF: Use separate cloudformation templates for gpu jobs (#299)
- ENH: Allow user to set number of GPUs for each job (#295)
- Use boto3 to get valid ec2 instance types (#284)
- Allow longer names for knots and fail earlier if they are too long (#282)
- ENH: Add nocache option to DockerImage.build() (#278)
- ENH: Add option to remove version pinning from pip requirements file (#277)
- FIX: Use default pickle protocol even for cloudpickle (#273)
- ENH: Add --ignore-installed option to ck.DockerImage (#275)
- ENH: Remove six (#274)
- FIX: Windows carriage return (#268)
- FIX: Allow user to specify name for new DockerImage instances (#265)
- ENH: Enforce maximum name length for Pars and Knot (#263)
- ENH: Automatically set job_type in knot.map() by inspecting the length of the arguments. (#261)
- FIX: Require object names to conform to AWS URI regex (#260)
- Add functions to clean up the cloudknot config file (#250)
- [ENH] Improve AWS resource tagging and allow user to provide custom tags (#249)
- Test Job Definition Name (#246)
- Remove config files for services we no longer use (#244)
- Create a better release message (#243)
- WIP: Fix zenodo metadata (#241)
- Use relative paths in docbuild.yml (#242)
- Allow user to refresh function for existing DockerImage (#240)
- Code style improvement suggestions from codacy (#237)
- Add publish_release script (#238)
- Add citation info to docs and update README badges (#236)
- Add back tags and use repo URI not image URI (#235)
- Adds some tools for maintenance. (#232)
- Update setup.py URL and bump version (#229)
- Fix volume_size and PARS policies issues (#233)