Feature
* feat: autopublish ([`a0900ff`](https://github.com/chanind/frame-semantic-transformer/commit/a0900ff546e99be008491b9b2446d9534e08af76))
Fix
* fix: pinning old version of semantic-release plugin ([`85f3a62`](https://github.com/chanind/frame-semantic-transformer/commit/85f3a625225da5c880b929c147bcf65bd629f2ea))
* fix: adding fetch-depth 0 for release ([`6dab4e6`](https://github.com/chanind/frame-semantic-transformer/commit/6dab4e661059cc8e895161cc07e3bec84719f784))
* fix: autopublish to pypi ([`3591c27`](https://github.com/chanind/frame-semantic-transformer/commit/3591c27a78ff7a687d378649b806dfacce00d775))
Unknown
* restrict model revisions ([`9a36fc4`](https://github.com/chanind/frame-semantic-transformer/commit/9a36fc419798422140826caefb21a1b355fa6b8c))
* adding explanation about lightning_logs dir ([`7898bba`](https://github.com/chanind/frame-semantic-transformer/commit/7898bbab8a6cbcbbb96c4d329bc4be64afcd1ae3))
* updating README and improving train script ([`94d7fac`](https://github.com/chanind/frame-semantic-transformer/commit/94d7fac13a3eae417cc6410f9ca973e568a5210a))
* Create LICENSE ([`f73fc0e`](https://github.com/chanind/frame-semantic-transformer/commit/f73fc0ea4720d960d499e92fb5fee07ed6f68579))
* augment training samples dynamically during training ([`3b40f07`](https://github.com/chanind/frame-semantic-transformer/commit/3b40f07f3538a7551a808220f24e7b89e5366bf5))
* adding tests for chain_augmentations ([`57f24dc`](https://github.com/chanind/frame-semantic-transformer/commit/57f24dc3ee559718dcb255e9ce2c752dd6522962))
* adding write permissions to publish job ([`d4b3e22`](https://github.com/chanind/frame-semantic-transformer/commit/d4b3e2264cc5e7b82d7a45c426b95beb00d854e3))
* try checkout v2 ([`7883323`](https://github.com/chanind/frame-semantic-transformer/commit/78833239d4735aaf52226c0d8d22be9d84dd229f))
* try adding token to checkout action ([`dce1651`](https://github.com/chanind/frame-semantic-transformer/commit/dce1651cdaf32b028f1c5e491e069c141da894a3))
* adding link to demo in README ([`f0b632b`](https://github.com/chanind/frame-semantic-transformer/commit/f0b632b09c921083910d93b8e2d3a2fd9ae3f9d6))
* fix node version in gh action ([`7db37cb`](https://github.com/chanind/frame-semantic-transformer/commit/7db37cb1f061ef460931a124127fefe3f864fbcd))
* add an action to publish the website ([`182504f`](https://github.com/chanind/frame-semantic-transformer/commit/182504f3dfb5049f6ed45030783dc5da5442fdd5))
* augment data for train but not val or test ([`660919e`](https://github.com/chanind/frame-semantic-transformer/commit/660919eba8a1637871bd712f1c3a7b033412ce40))
* adding data augmentation ([`13de3f4`](https://github.com/chanind/frame-semantic-transformer/commit/13de3f46d37deafe8bca94eb545b993d7ee71865))
* adding small size model and lazy-loading for the nltk + models ([`bda0ca8`](https://github.com/chanind/frame-semantic-transformer/commit/bda0ca81e9ed71d37ee3f508043e0d3d53bc1fc7))
* adding a demo client using create-react-app ([`2a673d1`](https://github.com/chanind/frame-semantic-transformer/commit/2a673d10f49ef4b444416489b723b679fb7c31af))
* try restricting batch size to 2 to avoid excessive memory use ([`4f2e58c`](https://github.com/chanind/frame-semantic-transformer/commit/4f2e58caffca5cd27b26b7b0d4b658afde62f9dd))
* try reducing to 1 thread to save memory ([`cfc4d21`](https://github.com/chanind/frame-semantic-transformer/commit/cfc4d216f13f2b78b037ad68f075a428253ad238))
* adding cors support to flask ([`4d463cb`](https://github.com/chanind/frame-semantic-transformer/commit/4d463cb453788b5149f7f498ff796d4c670045fa))
* increase gunicorn timeout ([`acd42a1`](https://github.com/chanind/frame-semantic-transformer/commit/acd42a14b89ad6aa34ee35affab1c2eb123ec83f))
* try adding poetry.lock to speed up docker build ([`86c9bcb`](https://github.com/chanind/frame-semantic-transformer/commit/86c9bcbb911ae6d9784806fa704e4f44e40bd401))
* bump to trigger cloud run build ([`c56e5b4`](https://github.com/chanind/frame-semantic-transformer/commit/c56e5b4145a8b7d8a2b43bb58aa3b3d668329d7c))
* bump to trigger cloud run build ([`a6d3094`](https://github.com/chanind/frame-semantic-transformer/commit/a6d30943222a3115baa3fa7ba98cce68b12aa2e2))
* remove poetry.lock from docker build ([`805adf4`](https://github.com/chanind/frame-semantic-transformer/commit/805adf4f5fee3e23219fd0f62684b4f8972ad6c5))
* adding a dockerizer flask server for demo purposes ([`8f192ac`](https://github.com/chanind/frame-semantic-transformer/commit/8f192acbf1058b582e343610b0f4f0c1eaf6d831))
* fixing typo ([`b7b9e34`](https://github.com/chanind/frame-semantic-transformer/commit/b7b9e34c232e6760f582aecf18a8b9b08a6cb8f8))
* adding a base FrameSemanticTransformer class to make it easy to parse sentences into frames ([`c0b78cf`](https://github.com/chanind/frame-semantic-transformer/commit/c0b78cfd5356e990302550f81dcbf4f75250727e))
* refactoring TaskSample classes into Tasks and TaskSamples ([`667f85e`](https://github.com/chanind/frame-semantic-transformer/commit/667f85ec8314e5d9922612904d99125fc568da51))
* fixing tests ([`eab2f96`](https://github.com/chanind/frame-semantic-transformer/commit/eab2f966268f08c95d951b824c8f60378c824b8e))
* more efficient loading of frame elements from framenet ([`4655768`](https://github.com/chanind/frame-semantic-transformer/commit/465576894ad3a7fda21eeca65c84fa3b7deabef5))
* add a check for invalid output in eval ([`0e9eb88`](https://github.com/chanind/frame-semantic-transformer/commit/0e9eb88ad70f103464acd47431dde1d5f52c0ae4))
* add a check for invalid output in eval ([`81a829a`](https://github.com/chanind/frame-semantic-transformer/commit/81a829a0593af001d9b4041e92532771909869d4))
* eval arg id similar to how sesame does it ([`73b2db9`](https://github.com/chanind/frame-semantic-transformer/commit/73b2db9bae939554d1a97f101b308eecbd707486))
* try adding in all possible frame elements into task intro for argument extraction ([`94e3c89`](https://github.com/chanind/frame-semantic-transformer/commit/94e3c89557fe0931ecc243e736d8d118e52318ac))
* updating frame id samples to be closer to how sesame does it ([`39376e5`](https://github.com/chanind/frame-semantic-transformer/commit/39376e548755cc32939c6097231f33cb854aad0d))
* fixing evaltuate function to work with batches predictions ([`7ca93d1`](https://github.com/chanind/frame-semantic-transformer/commit/7ca93d112eb35c61574dfb03f2e0052ef77a600f))
* force tranformers v4.18.0 to keep mypy happy ([`ac2c6a6`](https://github.com/chanind/frame-semantic-transformer/commit/ac2c6a65be536492099123c541b594bbf89e57f1))
* using multiple predictions when evaluating frame id task ([`d28961c`](https://github.com/chanind/frame-semantic-transformer/commit/d28961cd5f87d42e87d51219f293984d06ef1efc))
* fixing typo ([`197b93f`](https://github.com/chanind/frame-semantic-transformer/commit/197b93ffee1b38110ee0c3508966835c0e2d797d))
* trying to add eval into training ([`d22080b`](https://github.com/chanind/frame-semantic-transformer/commit/d22080bdc1801645aac4be6caad5e9062c37bbac))
* limiting task rebalancing ratio ([`451eb6c`](https://github.com/chanind/frame-semantic-transformer/commit/451eb6c0c7db51a0ad50d847f3270bd3283d13df))
* adding in task mix balancing ([`ecde0b2`](https://github.com/chanind/frame-semantic-transformer/commit/ecde0b2005c3f7a30698fa88528a3d41cd34e6cc))
* moving T5Tokenizer.cleanup into standardize_punct method ([`12ccf38`](https://github.com/chanind/frame-semantic-transformer/commit/12ccf38253566325718a226c776274175128a235))
* Trying out built-in clean up tokenization method ([`5f4a723`](https://github.com/chanind/frame-semantic-transformer/commit/5f4a723f187e6a21cba1201c71326089b426593c))
* allow tweaking predict params in eval ([`a549041`](https://github.com/chanind/frame-semantic-transformer/commit/a5490414a914ddc5b975b747ce4302eeacd83522))
* tweaking trigger processing to hopefully be more amenable to how the tokenizer works ([`2103f4d`](https://github.com/chanind/frame-semantic-transformer/commit/2103f4dd008b3e5e9856cbc19caac043f0e4578f))
* more readable eval print ([`726126d`](https://github.com/chanind/frame-semantic-transformer/commit/726126ddc977212a52963566de8fada7b48e188d))
* adding option to print eval failures ([`d23bf6f`](https://github.com/chanind/frame-semantic-transformer/commit/d23bf6f2fc2e7d0488f308ae1407d8f34ec862e0))
* adding a punct standardization step ([`d08877e`](https://github.com/chanind/frame-semantic-transformer/commit/d08877e0678c58abfe0d14e64c6673e68c03fb59))
* fixing linting ([`9250636`](https://github.com/chanind/frame-semantic-transformer/commit/9250636167d468ebd6df9aa6b20e8c9dbcfea495))
* tweaking frame id eval to match sesame logic ([`162f50f`](https://github.com/chanind/frame-semantic-transformer/commit/162f50fbcd26b20f0016312eab7cf0343c67393c))
* removing sample from dataloader, as it appears to break things ([`0e1dee0`](https://github.com/chanind/frame-semantic-transformer/commit/0e1dee00a9271b241a1f2d4446eaddc7e3cad407))
* fixing trigger samples and adding tests ([`6f19673`](https://github.com/chanind/frame-semantic-transformer/commit/6f1967352d01fc7305651ea57d73af689455d1f1))
* adding logging statements inside training function ([`a28a706`](https://github.com/chanind/frame-semantic-transformer/commit/a28a706238309f996bdb9ab3873d5bb4eb7f3fac))
* refactoring based on simple-t5 ([`d89e5db`](https://github.com/chanind/frame-semantic-transformer/commit/d89e5dbefd88db186ccbd0d1f70c42f65870226e))
* fixing evaluate typing ([`3cd3c33`](https://github.com/chanind/frame-semantic-transformer/commit/3cd3c33cf51434890783930f5a17470dd0d6ec52))
* fixing future annotations ([`16a227c`](https://github.com/chanind/frame-semantic-transformer/commit/16a227c846225c9f724de0ffc912118df5c61208))
* fixing bug in py 3.7 ([`ea953fc`](https://github.com/chanind/frame-semantic-transformer/commit/ea953fc1e4583b0ed7e41f7e1722cba1a4b3a9d1))
* refactoring and adding a target id task ([`37302bb`](https://github.com/chanind/frame-semantic-transformer/commit/37302bbfff7163df0199d8388e166c7b73f08bab))
* adding total to tqdm iteration ([`058a12c`](https://github.com/chanind/frame-semantic-transformer/commit/058a12c6db059322d2fee80c93a232a4ce5a26f1))
* fixing device issues ([`49a01f3`](https://github.com/chanind/frame-semantic-transformer/commit/49a01f32b9298c422289cc663f31e76aca5a327d))
* fixing typing ([`6012b09`](https://github.com/chanind/frame-semantic-transformer/commit/6012b092c355a4e79c74c474c30e884f2cd9d265))
* more efficient eval processing ([`4bbee5c`](https://github.com/chanind/frame-semantic-transformer/commit/4bbee5c813a5bb6daa3db7aca3c62b87637b1afd))
* add tqdm for eval progress ([`69d98ac`](https://github.com/chanind/frame-semantic-transformer/commit/69d98acb74b20e4b45571910c55f09ff8ce9c37a))
* tweaking evaluate ([`0f3db83`](https://github.com/chanind/frame-semantic-transformer/commit/0f3db83383f109e744905b3951cb6d8b46da99e0))
* adding evaluate / predict helpers ([`be08ec1`](https://github.com/chanind/frame-semantic-transformer/commit/be08ec19e9bc4638f379030621e818235f58354d))
* adding fulltext filenames from sesame for eval ([`081533b`](https://github.com/chanind/frame-semantic-transformer/commit/081533b78e3f42e0f899b995ffa9d7fc0ac1ed3e))
* removing validation loop end as well ([`1c5dfb2`](https://github.com/chanind/frame-semantic-transformer/commit/1c5dfb252bf7c6eb15eb839702dc75839abd2f35))
* removing return from training loop end ([`ace9621`](https://github.com/chanind/frame-semantic-transformer/commit/ace96214b5ad10a174d1a080b1a4a331050f86dc))
* adding in closure... ([`4b8e184`](https://github.com/chanind/frame-semantic-transformer/commit/4b8e18441b90c905b93637a9fab4a883784b4fb0))
* updating optimzier_step ([`ea79146`](https://github.com/chanind/frame-semantic-transformer/commit/ea791461efbbadc49e6abaef71b854a64b14e15b))
* fixing typo ([`7aa0a49`](https://github.com/chanind/frame-semantic-transformer/commit/7aa0a49ec4d170790be996a4fe4488e4bd2435fa))
* moving dataset generation out of the tuner ([`ac9f90f`](https://github.com/chanind/frame-semantic-transformer/commit/ac9f90f6be66e80ca15e24b6599716aaa3834378))
* adding future annotations stuff ([`611a1a5`](https://github.com/chanind/frame-semantic-transformer/commit/611a1a5eefc3b57b801e8f17f55b2ebe70d185ca))
* adding future annotations stuff ([`b6c8a53`](https://github.com/chanind/frame-semantic-transformer/commit/b6c8a53ca426badfb30bed495aea8c0954636cbc))
* setting up a model for training ([`abfc3d8`](https://github.com/chanind/frame-semantic-transformer/commit/abfc3d8a448675b2d6d1b752bbb30560dc26bdfd))
* skipping confusing frames for now ([`1888b8c`](https://github.com/chanind/frame-semantic-transformer/commit/1888b8cd5e9af73f70667cb37dc38d19e868ecb5))
* adding helper for parsing examples from docs ([`8856b2e`](https://github.com/chanind/frame-semantic-transformer/commit/8856b2edacf96c83c466baf9da9577e2e435a1a8))
* fixing mypy ([`14e4832`](https://github.com/chanind/frame-semantic-transformer/commit/14e4832b6b30e674369edbdd43c12634b8cce489))
* fixing black formatting ([`a24193a`](https://github.com/chanind/frame-semantic-transformer/commit/a24193ada5dba6d59cf5133840488184d87e3274))
* initial commit ([`4df6628`](https://github.com/chanind/frame-semantic-transformer/commit/4df6628f14781b1da9b2b9defecd27b1e37250f5))