v0.4rc0 (2018-11-10)
Full Changelog Implemented enhancements:
Fixed bugs:
- import ALWAYS wait state into top level module #201
- standalone non-deterministically drops jobs when > 50 instances are spun up #197
- Pickling pytorch model causes error #190
- Executor won't run from shard if numshards = 1 #189
- restore __version__ #186
- lapack/blas calls very slow on multicore machines for stand alone mode #137
Closed issues:
- Python 3.7 runtime #267
- LocalInvoker only on Linux breaks all CLI on OSX/(maybe windows) #291
- click 7.0 release changed command line options #288
- standalone setup improvements #284
- sklearn usage example #280
- wrenhandler imports fcntl which is system specific #272
- WaitTest::test_all_complete fails in standalone mode some times due to botocore timeouts #268
- pywren setup error: no module named pywren #261
- Bucket already exists error #260
- No module named 'pwd' #259
- pywren-setup: ModuleNotFoundError: No module named 'pwd' #257
- create_instance_profile fails if profile already exists #254
- eu-west-1 runtime bucket not existing / not public #252
- Question about map result and wait function #250
- Pywren and Amazon RDS issue #248
- Guide needed: How to use' extra_meta ' parameter in pywren executor? #244
- pywren not functioning as expected on windows develepment env #243
- pywren-setup ends with no module named pywren #242
- Provide profile option #240
- Add license headers to all files #234
- AccessDenied after installation with default settings? #232
- Multiple Executors on single standalone instance #216
- half-implemented wrapped stream handler causes runtime untar errors on py2.7 #211
- remove numpy dependency #205
- SQS Create queue ignores custom config file #204
- List keys with prefix assumes at least one entry #203
- Add download statistics back for func/data/output #188
- pywren.wait() only works for futures of same job #179
- Clean up devnotes #165
- Passing pylint should be a prerequisite for actual travis tests #149
Merged pull requests:
- Release0.4/new runtime buckets #293 (Vaishaal)
- fix for non-linux machines #292 (zehric)
- fix standalone setup errors #289 (zehric)
- fix cli options for click 7.0 #287 (zehric)
- added sqs queue size and instance uptimes in cli #285 (zehric)
- Fix typo in s3_backend.py #283 (rjboczar)
- clean up local invoker #282 (zehric)
- refactor wrenhandler to be less os-specific #281 (zehric)
- fix for windows path separator issues #279 (zehric)
- windows fix for file locking #278 (zehric)
- fix wait on multiple callset_ids #277 (zehric)
- windows fix for pwd #276 (zehric)
- add exp backoff in jobrunner get_object so it doesn't fail on start #269 (Vaishaal)
- newer versions of pylint became more strict #265 (Vaishaal)
- Custom lookup runtime regions #255 (ericmjonas)
- do not error if iam role already exists during setup #246 (Vaishaal)
- use items() not iteritems() since python3 doesn't have iteritems() #245 (Vaishaal)
- License update for serialize module #241 (ooq)
- Add license headers. #238 (ooq)
- Move jobrunner to dir #228 (ericmjonas)
- Standalone multiple executors #224 (Vaishaal)
- Process failure properly captured #223 (ericmjonas)
- Travis #213 (apengwin)
- move from using wrappedstreamhandler to bytesio and read #212 (ericmjonas)
- Fix two simple bugs #207 (ericmjonas)
- remove numpy dep #206 (ericmjonas)
- Fix issue 199 #200 (ericmjonas)
- Standalone bug-fixes #198 (ericmjonas)
- Add ability to cancel a job [WIP] #196 (ericmjonas)
- Created per-handler-type env vars to fix #137 #194 (ericmjonas)
- Fix issue 186 #193 (ericmjonas)
- add download stats back in, fixes 188 #192 (ericmjonas)
- Upgrade cloudpickle to 0.4.0 #191 (ericmjonas)
- Migrate s3io to jobrunner #180 (ericmjonas)
- Docs #168 (apengwin)
v0.3.0 (2017-09-07)
Fixed bugs:
- Test Function doesn't parse command-line config #166
Closed issues:
- Error with None result #183
- Create max list seatbelt #176
- Make ALL_COMPLETED, ANY_COMPLETED, and ALWAYS available without having to specifically import #174
- Generic Release clean-up #171
- Change default runtime URLS to be region-specific #164
- config files manually loaded must be patched with storage config #162
- Straggler due to list-after-write consistency #160
- Do not proceed with travis build if pylint fails. #157
- I can't run test function #156
- Describe .pywren_config in "getting started" #134
- Mapping over an empty list should return an empty list #129
- Create License File #116
- Refactor storage APIs #108
- Too big of runtime #105
- Update
getting startedwebpage on the interactive script #103 - Have setup.py check for supported python versions #101
- Support 3.4 runtime (and others) #95
- Migrate travis script to using sec-since-epoch as GUID #91
- Consider using boto3 s3 transfer interface #23
- ImportError: No module named 'wren' #19
Merged pull requests:
- [Issue#183] removing none result check #184 (ooq)
- Fix issue 91 #182 (ericmjonas)
- Release 0.3 #178 (ericmjonas)
- New max-limit seatbelt #177 (ericmjonas)
- Fix setup py version #173 (ericmjonas)
- Lambda toobig seatbelts #172 (ericmjonas)
- Issue#160 Straggler due to list-after-write consistency #170 (ooq)
- Switch to region specific runtimes #169 (ericmjonas)
- Fix patching of storage handler config, then fix commandline #167 (ericmjonas)
- change docstring #147 (apengwin)
- Enable Pylint checks for PyWren #144 (shivaram)
- Pylint cleanup #143 (shivaram)
- Remove cloudpickle dependency and use local cloudpickle. #142 (ooq)
- add exclude modules argument to map #140 (Vaishaal)
- Handle empty input list in map #133 (shivaram)
- Add Apache v2 License #132 (shivaram)
- [Issue#108] Refactor Storage API #119 (ooq)
- fixed minor spelling mistake #118 (sean-smith)
- Added validation for s3 bucket names and made username portion of def… #117 (sean-smith)
- Fix issue with bucket creation in us-east-1 #113 (Donohue)
v0.2 (2017-03-27)
Closed issues:
- Modules do not serialize in ipython notebook #84
- Create an example that uses wait() #67
- Migrate to github org #62
- use wait in the examples #58
- Add links to RISELab #45
- create pypi package #38
- Investigate using EMR as a better backend for execution for long-running jobs #12
- When creating the lambda we indiscriminately upload too much #98
- Standardize configuration for executors #86
- pywren not working for numba functions in 0.1 due to module serialization weirdness #83
- pywren not respecting non UTF-8 encoding (if coding is provided at top of file) #82
- Only *.py in modules are uploaded #79
- Handle exception pickling #78
- Capture exception traceback from remote #77
- Turn on runtime sharding by default #75
- Why do we use multiprocess instead of multiprocessing #60
- Generic refactor / split wren.py / run pyflakes #59
- Testing for pywrencli.py #55
- Better synchronization of runtime information with local client #44
- Create S3 bucket if it doesn't exist #35
- Add documentation on permissions #27
- Fix directories so that the conda command-line utils have right path #18
- Create interactive getting started script #16
Merged pull requests:
- external shared s3 client #115 (ericmjonas)
- Fix issue 98 #104 (ericmjonas)
- properly set path and test #100 (ericmjonas)
- Serialization bugs #99 (ericmjonas)
- Refactor uses of config in executor. #97 (shivaram)
- switch to multiprocessing #94 (ericmjonas)
- Use S3 paginators API. #93 (ooq)
- Bug fix in s3util. #92 (ooq)
- Split wren.py #90 (ooq)
- Interactive setup script (Issue #16) #89 (ericmjonas)
- Move runtime sharding into config variable #88 (shivaram)
- [Issue#77] Print out traceback for exceptions from wrenhandler.py. #87 (ooq)
- fix to issue 84, modules work in interactive mode #85 (Vaishaal)
- Release branch #81 (ericmjonas)
v0.1 (2017-03-07)
* This Change Log was automatically generated by github_changelog_generator