v0.5.0

What's new

Added 🎉

• Added TrainingEngine abstraction to torch integration.
• Added FairScale with a FairScaleTrainingEngine
  that leverages FairScale's FullyShardedDataParallel. This is meant to be used within the TorchTrainStep.
• All PyTorch components (such as learning rate schedulers, optimizers, data collators, etc) from the
  transformers library and now registered under the corresponding class in the torch integration.
  For example, transformers Adafactor optimizer is registered as an Optimizer under the name
  "transformers::Adafactor". More details can be found in the documentation for the transformers integration.

Changed ⚠️

• Various changes to the parameters othe TorchTrainStep due to the introduction of the TrainingEngine class.
• Params logged as DEBUG level instead of INFO to reduce noise in logs.
• The waiting message for FileLock is now clear about which file it's waiting for.
• Added an easier way to get the default Tango global config
• Most methods to TorchTrainCallback also take an epoch parameter now.
• WandbTrainCallback now logs peak GPU memory occupied by PyTorch tensors per worker. This is useful because W&B's system metrics only display the total GPU memory reserved by PyTorch, which is always higher than the actual amount of GPU memory occupied by tensors. So these new metrics give a more accurate view into how much memory your training job is actually using.
• Plain old Python functions can now be used in Lazy objects.
• LocalWorkspace now creates a symlink to the outputs of the latest run.
• Tango is now better at guessing when a step has died and should be re-run.
• Tango is now more lenient about registering the same class under the same name twice.
• When you use dict instead of Dict in your type annotations, you now get a legible error message. Same for List, Tuple, and Set.

Fixed ✅

• Fixed a bug in Registrable and FromParams where registered function constructors would not properly construct
  arguments that were classes.
• Fixed a bug in FromParams that would cause a crash when an argument to the constructor had the name params.
• Made FromParams more efficient by only trying to parse the params as a Step when it looks like it actually could be a step.
• Fixed bug where Executor would crash if git command could not be found.
• Fixed bug where validation settings were not interpreted the right way by the torch trainer.
• When you register the same name twice using Registrable, you get an error message. That error message now contains the correct class name.

Commits

a39a69f Merge pull request #161 from allenai/FromParamsDuJour
3063a92 CHANGELOG quick fix
cd006ae Add TrainEngine abstraction to TorchTrainStep, add FairScale integration, improve transformers integration (#77)
93438eb Update setuptools requirement from <=59.5.0 to <60.8.0 (#170)
e57dd91 Bump sphinx-copybutton from 0.4.0 to 0.5.0 (#174)
a8b1bdc split Docker build into seperate workflow, only run when necessary (#178)
59c91f7 make install comments work on all shells (#179)
a059416 Merge pull request #160 from allenai/GuessStepDirBetter
de7195d more fixes for conda-forge (#177)
75e9d42 use conda in Docker image, multi-stage build (#172)
611e446 Merge pull request #176 from allenai/latest-outputs
7241d20 Merge pull request #175 from allenai/self-contained-tests
83aa692 Merge pull request #153 from allenai/LazyWithoutFromParams
893e601 use virtualenv within Docker (#167)
178b8bd Merge pull request #171 from allenai/LenientRegister
6c765c8 Merge pull request #169 from allenai/InformativeFileLock
91ff7ac Merge pull request #168 from allenai/DefaultGlobalConfig
5d602fb push Docker images to GHCR.io (#166)
2b26fc8 set 'resume' to 'allow' instead of 'auto' (#155)
26771e7 fix bug when git missing (#163)
9009119 Add Dockerfile (#162)
a02155d Add a required flag to the README for gpt2-example (#159)

v0.4.0

What's new

Changed ⚠️

• Default log level is WARNING instead of ERROR.
• The web UI now renders the step graph left-to-right.
• The web UI now shows runs by date, with the most recent run at the top.
• The web UI now shows steps in a color-coded way.
• The --include-package flag now also accepts paths instead of module names.

Fixed ✅

• Ensure tqdm log lines always make it into the log file out.log even when log level is WARNING or ERROR.

Commits

5ff51d6 Fix GPT2 example. (#158)
4011482 make --include-package accept paths (#157)
92b8fe5 Merge pull request #148 from allenai/RunsWithDates
a4417e5 fix gpt2 config
797e3e8 minor logging tweaks (#145)
42654e6 Prepare for release v0.4.0rc5
df301ef Merge pull request #119 from allenai/RunGeneration
42535c9 Add TorchEvalStep to torch integration, use "device_count" in TorchTrainStep instead of "devices" (#120)
ecc6087 Store log output to a file in run directory, other logging improvements (#132)
9008255 Merge pull request #141 from allenai/dependabot/pip/sphinx-4.4.0
8e09b66 Merge pull request #139 from allenai/remove-no-logging
26318b2 Merge pull request #133 from allenai/dependabot/pip/mypy-0.931
7c91c4b Merge pull request #126 from allenai/dependabot/pip/sphinx-4.3.2
a178076 Merge pull request #128 from allenai/dependabot/pip/mypy-0.930
106a8cf Merge pull request #130 from allenai/dependabot/pip/furo-2022.1.2
6db7b4c Merge pull request #129 from allenai/run-name-at-end
b93d22d Add typing info to package (PEP 561) (#131)
e8867ae fix StopEarlyCallback state recovery
465a525 Prepare for release v0.4.0rc4
7a65540 CHANGELOG quick fix
31622d1 add logo to docs and README (#121)
b17d325 fix bug with StepInfo (#122)
d5698a0 Bump myst-parser from 0.16.0 to 0.16.1 (#118)
69d1dc8 Bump mypy from 0.910 to 0.920 (#117)
9739944 Prepare for release v0.4.0rc3
20138ce improve release notes generation script
760b4f2 Add DatasetsFormat, making LoadDataset cacheable, fix bug with KeyboardInterrupt (#114)
e51691f Improvements to W&B callback (#115)
d044f6e Add pre/post epoch callbacks (#113)
ae1ae0b Bump myst-parser from 0.15.2 to 0.16.0 (#111)
c605a1e Merge pull request #90 from allenai/SqliteDictFormat
8d6804d Prepare for release v0.4.0rc2
f404541 Merge pull request #110 from allenai/Conda
c154c92 Merge pull request #101 from allenai/Euler
288e02f ensure all integrations are imported if we can't find registered name (#109)
e033170 Bump black from 21.11b1 to 21.12b0 (#102)
b2781fd FAQ not FAQs :) (#108)
81e1225 add FAQs to docs (#107)
2148b74 Merge pull request #106 from allenai/Favicon
f812b73 fix bug with resolving lazy step inputs (#105)
110fc79 Merge pull request #104 from allenai/why-tango
eab62d8 fix bug in distributed training (#103)
901631f Removed scary warning
e86b843 Better summary
3bfd7c1 adjust dependency pinning (#100)
72aaa53 Merge pull request #79 from allenai/jon/html-viz
d45ec0c Merge pull request #99 from allenai/SkipGitTests
98b022b make sure workspaces can be imported from base module (#98)
07ac494 Merge pull request #97 from allenai/lower-click-pin
5ccefa9 fix prelease indicator condition in CI
04a5ab8 Prepare for release v0.4.0rc1
b2c09e3 fix typo in example (#96)
aba5758 Merge pull request #94 from allenai/dependabot/pip/cached-path-gte-0.3.3-and-lt-1.1.0
48b0b24 Merge pull request #92 from allenai/dependabot/pip/datasets-gte-1.12-and-lt-1.17
0107672 Merge pull request #91 from allenai/dependabot/pip/sphinx-4.3.1
7d9d919 Merge pull request #93 from allenai/NoOverrides
b5907de Merge pull request #67 from allenai/ResponsibleSteps
20951ea Bump furo from 2021.11.16 to 2021.11.23 (#89)
8bb00c4 Bump black from 21.11b0 to 21.11b1 (#88)
f240ac4 update filelock + cached_path, improve release scripts (#87)
eba4b8e Bump black from 21.10b0 to 21.11b0 (#86)
bc80bb8 Merge pull request #85 from allenai/dependabot/pip/filelock-gte-3.3-and-lt-3.5
17d28c7 Bump furo from 2021.11.15 to 2021.11.16 (#84)
4118912 Merge pull request #82 from allenai/dependabot/pip/furo-2021.11.15
cb1b853 Merge pull request #62 from allenai/dependabot/pip/pytorch-lightning-gte-1.4.0-and-lt-1.6.0
30f7a13 W&B log as step+1 (#76)
aab58b6 add some conda instructions to CONTRIBUTING.md (#81)

v0.4.0rc5

What's new

Added 🎉

• Added TorchEvalStep to torch integration, registered as "torch::eval".

Changed ⚠️

• Renamed aggregate_val_metric to auto_aggregate_val_metric in TorchTrainStep.
• devices parameter to TorchTrainStep replaced with device_count: int.
• Run name printed at the end of a run so it's easier to find.
• Type information added to package data. See PEP 561 for more information.
• A new integration, transformers, with two new steps for running seq2seq models.
• Added logging_tqdm, if you don't want a progress bar, but you still want to see progress in the logs.
• Added threaded_generator(), for wrapping generators so that they run in a separate thread from the generator's consumer.
• Added a new example for evaluating the T0 model on XSum, a summarization task.
• Added MappedSequence for functionally wrapping sequences.
• Added TextFormat, in case you want to store the output of your steps in raw text instead of JSON.
• Steps can now list arguments in SKIP_ID_ARGUMENTS to indicate that the argument should not affect a step's
  unique id. This is useful for arguments that affect the execution of a step, but not the output.
• Step now implements __str__, so steps look pretty in the debugger.
• Added DatasetCombineStep, a step that combines multiple datasets into one.
• Added common.logging.initialize_worker_logging() function for configuring logging from worker processes/threads.
• Logs from tango run ... will be written to a file called out.log in the run directory.

Fixed ✅

• Fixed torch StopEarlyCallback state not being recovered properly on restarts.
• Fixed file friendly logging by removing special styling characters.
• Ensured exceptions captured in logs.
• LocalWorkspace now works properly with uncacheable steps.
• When a Tango run got killed hard, with kill -9, or because the machine lost power, LocalWorkspace would
  sometimes keep a step marked as "running", preventing further executions. This still happens sometimes, but it
  is now much less likely (and Tango gives you instructions for how to fix it).
• To make all this happen, LocalWorkspace now saves step info in a Sqlite database. Unfortunately that means that
  the workspace format changes and existing workspace directories won't work properly with it.
• Fixed premature cleanup of temporary directories when using MemoryWorkspace

Commits

df301ef Merge pull request #119 from allenai/RunGeneration
42535c9 Add TorchEvalStep to torch integration, use "device_count" in TorchTrainStep instead of "devices" (#120)
ecc6087 Store log output to a file in run directory, other logging improvements (#132)
9008255 Merge pull request #141 from allenai/dependabot/pip/sphinx-4.4.0
8e09b66 Merge pull request #139 from allenai/remove-no-logging
26318b2 Merge pull request #133 from allenai/dependabot/pip/mypy-0.931
7c91c4b Merge pull request #126 from allenai/dependabot/pip/sphinx-4.3.2
a178076 Merge pull request #128 from allenai/dependabot/pip/mypy-0.930
106a8cf Merge pull request #130 from allenai/dependabot/pip/furo-2022.1.2
6db7b4c Merge pull request #129 from allenai/run-name-at-end
b93d22d Add typing info to package (PEP 561) (#131)
e8867ae fix StopEarlyCallback state recovery

v0.4.0rc4

What's new

Fixed ✅

• Fixed a bug where StepInfo fails to deserialize when error is an exception that can't be pickled.

Commits

7a65540 CHANGELOG quick fix
31622d1 add logo to docs and README (#121)
b17d325 fix bug with StepInfo (#122)
d5698a0 Bump myst-parser from 0.16.0 to 0.16.1 (#118)
69d1dc8 Bump mypy from 0.910 to 0.920 (#117)

v0.4.0rc3

What's new

Added 🎉

• Added DatasetsFormat format and LoadStreamingDataset step to datasets integration.
• SqliteDictFormat for datasets.
• Added pre_epoch() and post_epoch() callback methods to PyTorch TrainCallback.

Changed ⚠️

• LoadDataset step from datasets integration is now cacheable, using the DatasetsFormat format by default.
  But this only works with non-streaming datasets. For streaming datasets, you should use the LoadStreamingDataset step instead.

Fixed ✅

• Fixed bug where KeyboardInterrupt exceptions were not handled properly by steps and workspaces.
• WandbTrainCallback now will use part of the step's unique ID as the name for the W&B run by default, to make
  it easier to indentify which tango step corresponds to each run in W&B.
• WandbTrainCallback will save the entire TrainConfig object to the W&B config.

Commits

20138ce improve release notes generation script
760b4f2 Add DatasetsFormat, making LoadDataset cacheable, fix bug with KeyboardInterrupt (#114)
e51691f Improvements to W&B callback (#115)
d044f6e Add pre/post epoch callbacks (#113)
ae1ae0b Bump myst-parser from 0.15.2 to 0.16.0 (#111)
c605a1e Merge pull request #90 from allenai/SqliteDictFormat

v0.4.0rc2

What's new

Added 🎉

• Sample experiment configurations that prove Euler's identity

Changed ⚠️

• Loosened Click dependency to include v7.0.
• Loosened datasets dependency.
• Tightened petname dependency to exclude next major release for safety.

Fixed ✅

• Workspace, MemoryWorkspace, and LocalWorkspace can now be imported directly from the tango
  base module.
• Uncacheable leaf steps would never get executed. This is now fixed.
• We were treating failed steps as if they were completed by accident.
• The visualization had a problem with showing steps that never executed because a dependency failed.
• Fixed a bug where Lazy inputs to a Step would fail to resolve arguments that come from the result
  of another step.
• Fixed a bug in TorchTrainStep where some arguments for distributed training (devices, distributed_port) weren't being set properly.

Commits

f404541 Merge pull request #110 from allenai/Conda
4c347db Merge branch 'main' into Conda
c154c92 Merge pull request #101 from allenai/Euler
b3a8ae6 Revert "Make sure default steps are available when you run tango run"
76bda76 Merge remote-tracking branch 'origin/main' into Euler
e073925 Revert "Import integrations safely"
5550ff8 Adds Conda to the readme
288e02f ensure all integrations are imported if we can't find registered name (#109)
e033170 Bump black from 21.11b1 to 21.12b0 (#102)
2fd48cc Merge branch 'main' into Euler
b2781fd FAQ not FAQs :) (#108)
b31c8f2 Merge branch 'main' into Euler
81e1225 add FAQs to docs (#107)
dec4a86 Merge remote-tracking branch 'origin/Euler' into Euler
23d7824 Brings back the Euler example
4327e22 Merge branch 'main' into Euler
2148b74 Merge pull request #106 from allenai/Favicon
5ce5c64 Merge branch 'main' into Favicon
18e556a Revert "Changelog"
97dd79d Merge branch 'main' into Euler
f812b73 fix bug with resolving lazy step inputs (#105)
0a05047 Merge branch 'main' into Euler
89b315a Tango Favicon
110fc79 Merge pull request #104 from allenai/why-tango
a67c617 Update README.md
a5f5bd0 Merge branch 'main' into why-tango
eab62d8 fix bug in distributed training (#103)
e73a4a0 Update README.md
eb3ef41 Merge branch 'main' into Euler
c1c9bd0 Changelog
901631f Removed scary warning
e86b843 Better summary
cf9088d We no longer start the server during tests
b257e46 Import integrations safely
10ec4f4 Pick the right mutable mapping
a175370 Make sure default steps are available when you run tango run
1d4051a Make mypy happy
67e8a02 Moved complex arithmetic to https://github.com/allenai/tango-example
8a5c0d7 Adds test for steps that fail
3d551b3 Test for uncacheable leaf steps
21f0280 Don't explicitly run steps that don't need it
5fa4731 isort
9a3cd52 Changelog
6df6682 clarify comment
c46e0b2 also pick any uncacheable direct dependencies of leaf steps
c32a5cc Example steps for Euler's identity
1c97f10 Write step info for all steps before running any of them
3336a7e Reset step info when re-running a failed step
228b0f3 How did this slip by?
f9f1d05 Make sure to execute uncacheable steps that are not a dependency of anything
3d44720 Add steps for complex arithmetic
3bfd7c1 adjust dependency pinning (#100)
72aaa53 Merge pull request #79 from allenai/jon/html-viz
adea827 Merge branch 'jon/html-viz' of github.com:allenai/tango into jon/html-viz # Please enter a commit message to explain why this merge is necessary, # especially if it merges an updated upstream into a topic branch. # # Lines starting with '#' will be ignored, and an empty message aborts # the commit.
e771da7 make alert look nice
d21a172 Merge branch 'main' into jon/html-viz
d45ec0c Merge pull request #99 from allenai/SkipGitTests
28f17fa Don't check for git if we're not running in a repo
7db639e Merge branch 'main' into jon/html-viz
98b022b make sure workspaces can be imported from base module (#98)
07ac494 Merge pull request #97 from allenai/lower-click-pin
f10b178 Merge branch 'main' into lower-click-pin
5ccefa9 fix prelease indicator condition in CI
9843c5a update CHANGELOG
91d6551 Merge branch 'main' into lower-click-pin
58dfd3e CHANGELOG.md
8a069f0 loosen Click requirement
85b7417 Shows an ugly but functional brief popup when you copy something
626f1b6 Merge branch 'main' into jon/html-viz
b0cbf63 make tooltips meaningful since i cannot remove them
422ae0d Update tango/server/report.js
18869b7 Make local results work
3b1326a Python 3.7 again
5a282ff Warning to the future
2f54569 Give the option of tracking dependencies properly through Workspace
f9f7bfb We don't have this method anymore.
77ae786 Silence a warning
0d532c5 Reformat setup.py only in Python 3.7?
4f1e786 Changelog
b500f0e Merge branch 'main' into jon/html-viz
5c0b794 Merge branch 'main' into jon/html-viz
5c455e2 Include assets in package
4153e9e Fix import order
85dd1af Print a direct link to the run if we have one
01f0343 Makes it so you can run the server from any directory
300b722 Make frontend and backend consistent in their terminology
0ef1350 Fix some errors in the server
bd07ad5 Merge remote-tracking branch 'origin/ResponsibleSteps' into jon/html-viz
0cde4e7 pr updates
7bf9488 Makes tests pass
e96f6f3 Merge branch 'ResponsibleSteps' into jon/html-viz
8800bc9 Merge branch 'ResponsibleSteps' into jon/html-viz
378f649 pr gix
5047f6b add logo
4a7487b add in reloading every second
a1131ed pr fixes
95f1cbe pr fixes
d47b624 pr fixes
5962699 pr fixes
a974369 fix merge conflicts
3280491 initial code to serve and display tango viz

v0.4.0rc1

What's new

Added 🎉

• Introduced the concept of the Workspace, with LocalWorkspace and MemoryWorkspace as initial implementations.
• Added a stub of a webserver that will be able to visualize runs as they happen.
• Added separate classes for LightningTrainingTypePlugin, LightningPrecisionPlugin, LightningClusterEnvironmentPlugin, LightningCheckpointPlugin for compatibility with pytorch-lightning>=1.5.0.

Removed 👋

• Removed old LightningPlugin class
• Removed requirement of the overrides package

Changed ⚠️

• Made it possible to construct a step graph out of Step objects, instead of constructing it out of StepStub objects.
• Removed dataset fingerprinting code, since we can now use Step to make sure things are cached.
• Made steps deterministic by default.
• Brought back MemoryStepCache, so we can run steps without configuring anything.
• W&B torch::TrainCallback logs with step=step+1 now so that training curves in the W&B dashboard
  match up with checkpoints saved locally and are easier to read (e.g. step 10000 instead of 9999).
• filelock >= 3.4 required, parameter poll_intervall to tango.common.file_lock.FileLock.acquire renamed
  to poll_interval.

Fixed ✅

• Fixed bug in FromParams where a parameter to a FromParams class may not be instantiated correctly
  if it's a class with a generic type parameter.

Commits

b2c09e3 fix typo in example (#96)
aba5758 Merge pull request #94 from allenai/dependabot/pip/cached-path-gte-0.3.3-and-lt-1.1.0
4ae6115 Update requirements.txt
d4d0655 Merge branch 'main' into dependabot/pip/cached-path-gte-0.3.3-and-lt-1.1.0
48b0b24 Merge pull request #92 from allenai/dependabot/pip/datasets-gte-1.12-and-lt-1.17
6044f8b Update cached-path requirement from <0.4.0,>=0.3.3 to >=0.3.3,<1.1.0
1ae82aa Merge branch 'main' into dependabot/pip/datasets-gte-1.12-and-lt-1.17
0107672 Merge pull request #91 from allenai/dependabot/pip/sphinx-4.3.1
fa47e54 Merge branch 'main' into dependabot/pip/datasets-gte-1.12-and-lt-1.17
38c0b42 Merge branch 'main' into dependabot/pip/sphinx-4.3.1
7d9d919 Merge pull request #93 from allenai/NoOverrides
dfa461c Removes the dependency on the overrides package
ff18197 Update datasets requirement from <1.16,>=1.12 to >=1.12,<1.17
fc69dd0 Bump sphinx from 4.3.0 to 4.3.1
b5907de Merge pull request #67 from allenai/ResponsibleSteps
267a6e4 clean up config usage
82862ef Merge branch 'main' into ResponsibleSteps
20951ea Bump furo from 2021.11.16 to 2021.11.23 (#89)
8d8670a Optional server
03049fa Handle the log level consistently
c620405 Merge branch 'main' into ResponsibleSteps
8bb00c4 Bump black from 21.11b0 to 21.11b1 (#88)
cd5a70a Fix tests
71cfbd7 Don't cache uncacheable steps
7839fd8 Merge branch 'main' into ResponsibleSteps
f240ac4 update filelock + cached_path, improve release scripts (#87)
967ecb2 Merge branch 'main' into ResponsibleSteps
10634aa Don't show inherited from_params
4b104be Fix test
4e9910d Avoid a naming conflict in computer science
6db2c29 Improved documentation
44f79ec Added blurb
7f68a97 Use enum for step states
19e6de2 Click logging is disabled by default, enabled in the CLI use case
eba4b8e Bump black from 21.10b0 to 21.11b0 (#86)
1d80766 Log the start of a run
6190e2d Merge branch 'main' into ResponsibleSteps
3550c3f Cleaner workspace docs
08d4056 Better StepCache docs
d5ede4e Formatting
bc80bb8 Merge pull request #85 from allenai/dependabot/pip/filelock-gte-3.3-and-lt-3.5
d69bac4 Check whether a run name already exists
cde0f14 Unused import
a80a87f Fix the case where a step's cacheability changes across restarts
de5f248 Improve comment
f13d717 Losely pin petname
5d72923 Merge branch 'main' into ResponsibleSteps
914902b Merge branch 'ResponsibleSteps' of https://github.com/allenai/tango into ResponsibleSteps
2f5a266 Merge pull request #83 from allenai/petew-ResponsibleSteps
b3df88f Update filelock requirement from <3.4,>=3.3 to >=3.3,<3.5
17d28c7 Bump furo from 2021.11.15 to 2021.11.16 (#84)
8f2b48e add failing test case
1fc9860 Merge remote-tracking branch 'origin/main' into ResponsibleSteps
4118912 Merge pull request #82 from allenai/dependabot/pip/furo-2021.11.15
d5eb968 Merge branch 'main' into dependabot/pip/furo-2021.11.15
abac28a Fix the shortcut for running all (many) checks
c4009de Merge branch 'main' into ResponsibleSteps
e364766 Merge pull request #73 from allenai/petew-ResponsibleSteps
8ae0e27 More doctests
6211025 Format docs better
5814724 Fix docs
cb1b853 Merge pull request #62 from allenai/dependabot/pip/pytorch-lightning-gte-1.4.0-and-lt-1.6.0
50caada Update requirements.txt
9c79b97 Makes the docs build
3ecb952 Merge branch 'main' into dependabot/pip/pytorch-lightning-gte-1.4.0-and-lt-1.6.0
55c5fae Bump furo from 2021.11.12 to 2021.11.15
83add7e Important fixes
301347d Fix tests
cfe29de Create workdir when requested
7b67142 Bring back "needed by"
a46b3e8 This wasn't meant to be checked in.
af4cc3c Use click through the logger
ee81446 Merge branch 'ResponsibleSteps' into petew-ResponsibleSteps
7247528 We don't need this TODO right now.
9b8446c Merge branch 'main' into ResponsibleSteps
30f7a13 W&B log as step+1 (#76)
eab2b7f Merge pull request #74 from allenai/Workspaces
d91302b Simplify!
e605399 changelog
9f360da Bring back deterministic step randomness, without breaking random step names
a2c5cd8 fix order of imports
02ad624 Merge branch 'dependabot/pip/pytorch-lightning-gte-1.4.0-and-lt-1.6.0' of https://github.com/allenai/tango into dependabot/pip/pytorch-lightning-gte-1.4.0-and-lt-1.6.0
7f78b0a remove comment
8f8cbeb separating different plugin types
b139181 Merge branch 'main' into dependabot/pip/pytorch-lightning-gte-1.4.0-and-lt-1.6.0
761f873 Merge branch 'ResponsibleSteps' into Workspaces
12a480a Merge branch 'ResponsibleSteps' into petew-ResponsibleSteps
b7d4e88 merge main
63f6029 Merge branch 'main' into ResponsibleSteps
aab58b6 add some conda instructions to CONTRIBUTING.md (#81)
e7e5c5e Fix symlink creation
b547472 Adds a command to keep a server running permanently
596c278 Not sure how this line got lost
06d7681 Fixes and cleanup
3cebc24 fix bug caused by random seed
d145216 fix comment
067436d clean up
a4ce577 Creates and uses the concept of a workspace, so that the server can consume it
110eb07 ci
e811f47 fix tests
460dd87 executor fixes
837a454 handle generic non-FromParams classes
7fed5a6 fix merge conflicts
12ae8e9 Fixing the torch test
0c208e1 Remove stale comment
b44a0a0 Fix doctest
3e4eb42 Makes det_hash consistent across Python versions
f23b56c Executable documentation!
38388ca Formatting
6aaa01e Fix some documentation
57825a0 Fix docs
cb7591d Order imports correctly 🙄
8caea06 Removing unused imports
f7f5d2e Changelog again
05ca4dd Changelog
da16b5a Merge branch 'main' into ResponsibleSteps
8c51b45 Make nested steps work for classes that aren't FromParams
083516b Remember which extra modules we imported
54e7dbc We can't restore the registry like that.
455b756 Refactors the test to fail in new and exciting ways
62cd1d2 Merge branch 'main' into dependabot/pip/pytorch-lightning-gte-1.4.0-and-lt-1.6.0
cd68752 Don't need this comment anymore
76c10f1 Remove fingerprint stuff
a1f1af8 Merge branch 'FixImport' into ResponsibleSteps
ed9ebc9 Fix Import
14bc2ee Merge branch 'main' into ResponsibleSteps
4a4ab58 Merge branch 'main' into dependabot/pip/pytorch-lightning-gte-1.4.0-and-lt-1.6.0
e5ea1af Fix after merge
7b92dd9 Merge remote-tracking branch 'origin/main' into ResponsibleSteps
d066768 We were not actually using this function
df07690 Mypy inspired changes
87ee96f Formatting
3823f98 🤦🏼
d018838 Fix executor test
a7fb4ea Fix executor
0b1f310 Quiet, you
449a70d Fixes circular references
4fc1f34 Actually write the circular reference test
0df3538 Throw the error for the right reason
afb3b61 WithUnresolvedSteps
1be6816 Update pytorch-lightning requirement
3c4cb19 Dicts are iterable, so these have to be swapped
efc84d4 Adds new failing test
1f2bf16 Relative imports don't work
430112f Fix bug in test
c3dd08d Add test that fails
118f8c2 Better name for the test
a6ac3c9 Detect unsatisfiable dependencies
8eb7a48 Bring back parsing everything as a Step first :-/
bc6f886 Type checks
00b2df6 Makes the test pass
a6ba11a Make the code more compatible with the IDE
a611d48 We don't need these anymore
132153d Formatting
d019499 Start fixing hard tests
f4d77c1 Fix trivial tests
fa6de92 Make steps responsible for their own execution
cfe7007 Slightly more readable error message
c0ded32 Typo

v0.3.6

What's new

Added 🎉

• Added a .log_batch() method on torch::TrainCallback which is given the average loss across
  distributed workers, but only called every log_every steps.

Removed 👋

• Removed .pre_log_batch() method on torch::TrainCallback.

Fixed ✅

• Fixed typo in parameter name remove_stale_checkpoints in TorchTrainStep (previously was remove_state_checkpoints).
• Fixed bug in FromParams that would cause failures when from __future__ import annotations
  was used with Python older than 3.10. See PEP 563
  for details.

Commits

6b5cb24 support for PEP 563 in older Python versions (#80)
27657d9 Bump furo from 2021.10.9 to 2021.11.12 (#78)
c8ac858 Bump sphinx from 4.2.0 to 4.3.0 (#75)
5516244 refactor TorchTrainStep (#70)
b13bba7 Bump isort from 5.10.0 to 5.10.1 (#71)

v0.3.5

What's new

Fixed ✅

• Fixed a bug in FromParams where the "type" parameter was ignored in some cases
  where the Registrable base class did not directly inherit from Registrable.

Commits

5bdad24 Merge pull request #69 from allenai/weird-mix-fix
5434e23 test case
e1108a5 CHANGELOG.md
2bbfa6f fix weird FromParams/Registrable bug

v0.3.4

What's new

Added 🎉

• Added StopEarlyCallback, a TorchTrainCallback for early stopping.
• Added parameter remove_stale_checkpoints to TorchTrainStep.

Changed ⚠️

• Minor changes to TorchTrainCallback interface.
• Weights & Biases TorchTrainCallback now logs best validation metric score.

Commits

c126442 torch train improvements (#64)