Merge branch 'master' into feature/add-deepspeed-finetuning-strategies

.github/CODEOWNERS

@@ -10,20 +10,18 @@
 # owners
 /.github/CODEOWNERS   @williamfalcon
 # main
-/README.md            @edenlightning @ethanwharris
+/README.md            @ethanwharris @krshrimali
 # installation
-/setup.py                 @borda @ethanwharris
-/__about__.py             @borda @ethanwharris
-/__init__.py              @borda @ethanwharris
+/setup.py                 @borda @ethanwharris @krshrimali
+/__about__.py             @borda @ethanwharris @krshrimali
+/__init__.py              @borda @ethanwharris @krshrimali
 
 # CI/CD
-/.github/workflows/   @borda @ethanwharris
+/.github/workflows/   @borda @ethanwharris @krshrimali
 # configs in root
-/*.yml                @borda @ethanwharris
+/*.yml                @borda @ethanwharris @krshrimali
 
 # Docs
-/docs/                          @edenlightning @ethanwharris
-/.github/*.md                   @edenlightning @ethanwharris
-/.github/ISSUE_TEMPLATE/*.md    @edenlightning @ethanwharris
+/.github/ISSUE_TEMPLATE/*.md    @borda @ethanwharris @krshrimali
 /docs/source/conf.py            @borda @ethanwharris
 /flash/core/integrations/labelstudio	@KonstantinKorotaev @niklub

.github/labeler.yml

@@ -4,7 +4,6 @@ documentation:
 
 examples:
   - flash_examples/**/*
-  - flash_notebooks/**/*
 
 data:
   - flash/core/data/**/*

.gitignore

@@ -144,8 +144,6 @@ titanic.csv
 data_folder
 *.pt
 *.zip
-flash_notebooks/*.py
-flash_notebooks/data
 /data
 MNIST*
 titanic

CHANGELOG.md

@@ -10,6 +10,8 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 
 - Added fine tuning strategies for DeepSpeed (with parameter loading and storing omitted) ([#1377](https://github.com/Lightning-AI/lightning-flash/pull/1377))
 
+- Added `torchvision` as a requirement to `datatype_audio.txt` as it's used for Audio Classification ([#1425](https://github.com/Lightning-AI/lightning-flash/pull/1425))
+
 - Added `figsize` and `limit_nb_samples` for showing batch images ([#1381](https://github.com/Lightning-AI/lightning-flash/pull/1381))
 
 - Added support for `from_lists` for Tabular Classification and Regression ([#1337](https://github.com/PyTorchLightning/lightning-flash/pull/1337))
@@ -48,6 +50,8 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 
 ### Fixed
 
+- Fixed when suitable error not being raised for image segmentation (kornia) ([#1425](https://github.com/Lightning-AI/lightning-flash/pull/1425)).
+
 - Fixed the script of integrating `lightning-flash` with `learn2learn` ([#1376](https://github.com/Lightning-AI/lightning-flash/pull/1383))
 
 - Fixed JIT tracing tests where the model class was not attached to the `Trainer` class ([#1410](https://github.com/Lightning-AI/lightning-flash/pull/1410))

README.md

@@ -129,7 +129,7 @@ model.serve()
 or make predictions from raw data directly.
 
 ```py
-trainer = Trainer(accelerator='ddp', gpus=2)
+trainer = Trainer(strategy='ddp', accelerator="gpu", gpus=2)
 dm = SemanticSegmentationData.from_folders(predict_folder="data/CameraRGB")
 predictions = trainer.predict(model, dm)
 ```

docs/source/_templates/theme_variables.jinja

@@ -6,7 +6,7 @@
   'docs': 'https://lightning-flash.readthedocs.io',
   'twitter': 'https://twitter.com/PyTorchLightnin',
   'discuss': 'https://pytorch-lightning.slack.com',
-  'tutorials': 'https://github.com/PyTorchLightning/lightning-flash/tree/master/flash_notebooks',
+  'tutorials': 'https://github.com/Lightning-AI/tutorials',
   'previous_pytorch_versions': 'https://lightning-flash.readthedocs.io/en/stable',
   'home': 'https://lightning-flash.readthedocs.io',
   'get_started': 'https://lightning-flash.readthedocs.io/en/latest/quickstart.html',

docs/source/governance.rst

@@ -6,13 +6,13 @@ Flash Governance | Persons of interest
 Leads
 -----
 - Ethan Harris (`ethanwharris <https://github.com/ethanwharris>`_)
+- Kushashwa Ravi Shrimali (`krshrimali <https://github.com/krshrimali>`_)
 - Thomas Chaton (`tchaton <https://github.com/tchaton>`_)
-- William Falcon (`williamFalcon <https://github.com/williamFalcon>`_)
 
 Core Maintainers
 ----------------
+- William Falcon (`williamFalcon <https://github.com/williamFalcon>`_)
 - Jirka Borovec (`Borda <https://github.com/Borda>`_)
-- Kushashwa Ravi Shrimali (`krshrimali` <https://github.com/krshrimali>`_)
 - Kaushik Bokka (`kaushikb11 <https://github.com/kaushikb11>`_)
 - Justus Schock (`justusschock <https://github.com/justusschock>`_)
 - Akihiro Nitta (`akihironitta <https://github.com/akihironitta>`_)

flash/audio/classification/input_transform.py

@@ -19,7 +19,7 @@
 from flash.core.data.io.input import DataKeys
 from flash.core.data.io.input_transform import InputTransform
 from flash.core.data.transforms import ApplyToKeys
-from flash.core.utilities.imports import _TORCHAUDIO_AVAILABLE, _TORCHVISION_AVAILABLE
+from flash.core.utilities.imports import _TORCHAUDIO_AVAILABLE, _TORCHVISION_AVAILABLE, requires
 
 if _TORCHVISION_AVAILABLE:
     from torchvision import transforms as T
@@ -51,6 +51,7 @@ def train_per_sample_transform(self) -> Callable:
             ]
         )
 
+    @requires("audio")
     def per_sample_transform(self) -> Callable:
         return T.Compose(
             [

flash/core/data/io/input.py

@@ -15,6 +15,7 @@
 import os
 import sys
 from copy import deepcopy
+from enum import Enum
 from typing import Any, cast, Dict, Iterable, List, Sequence, Tuple, Union
 
 from pytorch_lightning.utilities.enums import LightningEnum
@@ -171,7 +172,18 @@ def __init__(self, running_stage: RunningStage, *args: Any, **kwargs: Any) -> No
 
     def _call_load_sample(self, sample: Any) -> Any:
         # Deepcopy the sample to avoid leaks with complex data structures
-        return getattr(self, f"{_STAGES_PREFIX[self.running_stage]}_load_sample")(deepcopy(sample))
+        sample_output = getattr(self, f"{_STAGES_PREFIX[self.running_stage]}_load_sample")(deepcopy(sample))
+
+        # Change DataKeys Enum to strings
+        if isinstance(sample_output, dict):
+            output_dict = {}
+            for key, val in sample_output.items():
+                if isinstance(key, Enum) and hasattr(key, "value"):
+                    output_dict[key.value] = val
+                else:
+                    output_dict[key] = val
+            return output_dict
+        return sample_output
 
     @staticmethod
     def load_data(*args: Any, **kwargs: Any) -> Union[Sequence, Iterable]:

flash/core/utilities/imports.py

@@ -163,7 +163,7 @@ class Image:
 )
 _SERVE_AVAILABLE = _FASTAPI_AVAILABLE and _PYDANTIC_AVAILABLE and _CYTOOLZ_AVAILABLE and _UVICORN_AVAILABLE
 _POINTCLOUD_AVAILABLE = _OPEN3D_AVAILABLE and _TORCHVISION_AVAILABLE
-_AUDIO_AVAILABLE = all([_TORCHAUDIO_AVAILABLE, _LIBROSA_AVAILABLE, _TRANSFORMERS_AVAILABLE])
+_AUDIO_AVAILABLE = all([_TORCHAUDIO_AVAILABLE, _TORCHVISION_AVAILABLE, _LIBROSA_AVAILABLE, _TRANSFORMERS_AVAILABLE])
 _GRAPH_AVAILABLE = (
     _TORCH_SCATTER_AVAILABLE and _TORCH_SPARSE_AVAILABLE and _TORCH_GEOMETRIC_AVAILABLE and _NETWORKX_AVAILABLE
 )

flash/image/segmentation/input_transform.py

@@ -17,7 +17,7 @@
 from flash.core.data.io.input import DataKeys
 from flash.core.data.io.input_transform import InputTransform
 from flash.core.data.transforms import ApplyToKeys, kornia_collate, KorniaParallelTransforms
-from flash.core.utilities.imports import _KORNIA_AVAILABLE, _TORCHVISION_AVAILABLE
+from flash.core.utilities.imports import _KORNIA_AVAILABLE, _TORCHVISION_AVAILABLE, requires
 
 if _KORNIA_AVAILABLE:
     import kornia as K
@@ -47,6 +47,7 @@ class SemanticSegmentationInputTransform(InputTransform):
     mean: Union[float, Tuple[float, float, float]] = (0.485, 0.456, 0.406)
     std: Union[float, Tuple[float, float, float]] = (0.229, 0.224, 0.225)
 
+    @requires("image")
     def train_per_sample_transform(self) -> Callable:
         return T.Compose(
             [
@@ -61,6 +62,7 @@ def train_per_sample_transform(self) -> Callable:
             ]
         )
 
+    @requires("image")
     def per_sample_transform(self) -> Callable:
         return T.Compose(
             [
@@ -72,6 +74,7 @@ def per_sample_transform(self) -> Callable:
             ]
         )
 
+    @requires("image")
     def predict_per_sample_transform(self) -> Callable:
         return ApplyToKeys(
             DataKeys.INPUT,

flash_examples/integrations/learn2learn/image_classification_imagenette_mini.py

@@ -146,6 +146,7 @@ def collate(self):
 trainer = flash.Trainer(
     max_epochs=1,
     gpus=1,
+    accelerator="gpu",
     precision=16,
 )