Lightning-flash

Latest version: v0.8.2

Safety actively analyzes 638763 Python packages for vulnerabilities to keep your Python projects secure.

Scan your dependencies

Page 3 of 4

0.4.0

Added

- Added integration with FiftyOne ([360](https://github.com/PyTorchLightning/lightning-flash/pull/360))
- Added `flash.serve` ([399](https://github.com/PyTorchLightning/lightning-flash/pull/399))
- Added support for `torch.jit` to tasks where possible and documented task JIT compatibility ([389](https://github.com/PyTorchLightning/lightning-flash/pull/389))
- Added option to provide a `Sampler` to the `DataModule` to use when creating a `DataLoader` ([390](https://github.com/PyTorchLightning/lightning-flash/pull/390))
- Added support for multi-label text classification and toxic comments example ([401](https://github.com/PyTorchLightning/lightning-flash/pull/401))
- Added a sanity checking feature to flash.serve ([423](https://github.com/PyTorchLightning/lightning-flash/pull/423))

Changed

- Split `backbone` argument to `SemanticSegmentation` into `backbone` and `head` arguments ([412](https://github.com/PyTorchLightning/lightning-flash/pull/412))

Fixed

- Fixed a bug where the `DefaultDataKeys.METADATA` couldn't be a dict ([393](https://github.com/PyTorchLightning/lightning-flash/pull/393))
- Fixed a bug where the `SemanticSegmentation` task would not work as expected with finetuning callbacks ([412](https://github.com/PyTorchLightning/lightning-flash/pull/412))
- Fixed a bug where predict batches could not be visualized with `ImageClassificationData` ([438](https://github.com/PyTorchLightning/lightning-flash/pull/438))

0.3.2

Fixed

- Fixed a bug where `flash.Trainer.from_argparse_args` + `finetune` would not work ([382](https://github.com/PyTorchLightning/lightning-flash/pull/382))

0.3.1

Added

- Added `deeplabv3`, `lraspp`, and `unet` backbones for the `SemanticSegmentation` task ([370](https://github.com/PyTorchLightning/lightning-flash/pull/370))

Changed

- Changed the installation command for extra features ([346](https://github.com/PyTorchLightning/lightning-flash/pull/346))
- Change resize interpolation default mode to nearest ([352](https://github.com/PyTorchLightning/lightning-flash/pull/352))

Deprecated

- Deprecated `SemanticSegmentation` backbone names `torchvision/fcn_resnet50` and `torchvision/fcn_resnet101`, use `fc_resnet50` and `fcn_resnet101` instead ([370](https://github.com/PyTorchLightning/lightning-flash/pull/370))

Fixed

- Fixed `flash.Trainer.add_argparse_args` not adding any arguments ([343](https://github.com/PyTorchLightning/lightning-flash/pull/343))
- Fixed a bug where the translation task wasn't decoding tokens properly ([332](https://github.com/PyTorchLightning/lightning-flash/pull/332))
- Fixed a bug where huggingface tokenizers were sometimes being pickled ([332](https://github.com/PyTorchLightning/lightning-flash/pull/332))
- Fixed issue with `KorniaParallelTransforms` to assure to share the random state between transforms ([351](https://github.com/PyTorchLightning/lightning-flash/pull/351))
- Fixed a bug where using `val_split` with `overfit_batches` would give an infinite recursion ([375](https://github.com/PyTorchLightning/lightning-flash/pull/375))
- Fixed a bug where some timm models were mistakenly given a `global_pool` argument ([377](https://github.com/PyTorchLightning/lightning-flash/pull/377))
- Fixed `flash.Trainer.from_argparse_args` not passing arguments correctly ([380](https://github.com/PyTorchLightning/lightning-flash/pull/380))

0.3.0

Added

- Added DataPipeline API ([188](https://github.com/PyTorchLightning/lightning-flash/pull/188) [#141](https://github.com/PyTorchLightning/lightning-flash/pull/141) [#207](https://github.com/PyTorchLightning/lightning-flash/pull/207))
- Added timm integration ([196](https://github.com/PyTorchLightning/lightning-flash/pull/196))
- Added BaseViz Callback ([201](https://github.com/PyTorchLightning/lightning-flash/pull/201))
- Added backbone API ([204](https://github.com/PyTorchLightning/lightning-flash/pull/204))
- Added support for Iterable auto dataset ([227](https://github.com/PyTorchLightning/lightning-flash/pull/227))
- Added multi label support ([230](https://github.com/PyTorchLightning/lightning-flash/pull/230))
- Added support for schedulers ([232](https://github.com/PyTorchLightning/lightning-flash/pull/232))
- Added visualisation callback for image classification ([228](https://github.com/PyTorchLightning/lightning-flash/pull/228))
- Added Video Classification task ([216](https://github.com/PyTorchLightning/lightning-flash/pull/216))
- Added Dino backbone for image classification ([259](https://github.com/PyTorchLightning/lightning-flash/pull/259))
- Added Data Sources API ([256](https://github.com/PyTorchLightning/lightning-flash/pull/256) [#264](https://github.com/PyTorchLightning/lightning-flash/pull/264) [#272](https://github.com/PyTorchLightning/lightning-flash/pull/272))
- Refactor preprocess_cls to preprocess, add Serializer, add DataPipelineState ([229](https://github.com/PyTorchLightning/lightning-flash/pull/229))
- Added Semantic Segmentation task ([239](https://github.com/PyTorchLightning/lightning-flash/pull/239) [#287](https://github.com/PyTorchLightning/lightning-flash/pull/287) [#290](https://github.com/PyTorchLightning/lightning-flash/pull/290))
- Added Object detection prediction example ([283](https://github.com/PyTorchLightning/lightning-flash/pull/283))
- Added Style Transfer task and accompanying finetuning and prediction examples ([262](https://github.com/PyTorchLightning/lightning-flash/pull/262))
- Added a Template task and tutorials showing how to contribute a task to flash ([306](https://github.com/PyTorchLightning/lightning-flash/pull/306))

Changed

- Rename valid_ to val_ ([197](https://github.com/PyTorchLightning/lightning-flash/pull/197))
- Refactor preprocess_cls to preprocess, add Serializer, add DataPipelineState ([229](https://github.com/PyTorchLightning/lightning-flash/pull/229))

Fixed

- Fix DataPipeline resolution in Task ([212](https://github.com/PyTorchLightning/lightning-flash/pull/212))
- Fixed a bug where the backbone used in summarization was not correctly passed to the postprocess ([296](https://github.com/PyTorchLightning/lightning-flash/pull/296))

0.2.3

Added

- Added TIMM integration as backbones ([196](https://github.com/PyTorchLightning/lightning-flash/pull/196))

Fixed

- Fixed nltk.download ([210](https://github.com/PyTorchLightning/lightning-flash/pull/196))

0.2.2

Changed

- Switch to use `torchmetrics` ([169](https://github.com/PyTorchLightning/lightning-flash/pull/169))

- Better support for `optimizer` and `schedulers` ([232](https://github.com/PyTorchLightning/lightning-flash/pull/232))

- Update lightning version to v1.2 ([133](https://github.com/PyTorchLightning/lightning-flash/pull/133))

Fixed

- Fixed classification softmax ([169](https://github.com/PyTorchLightning/lightning-flash/pull/169))

- Fixed a bug where loading from a local checkpoint that had `pretrained=True` without an internet connection would sometimes raise an error ([237](https://github.com/PyTorchLightning/lightning-flash/pull/237))

- Don't download data if exists ([157](https://github.com/PyTorchLightning/lightning-flash/pull/157))

Page 3 of 4

© 2024 Safety CLI Cybersecurity Inc. All Rights Reserved.