Merge branch 'master' into feat/detect

.github/mergify.yml

@@ -0,0 +1,58 @@
+# Copyright The PyTorch Lightning team.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+pull_request_rules:
+
+  - name: warn on conflicts
+    conditions:
+      - conflict
+      - -draft # filter-out GH draft PRs
+      - -label="has conflicts"
+    actions:
+      # comment:
+      #   message: This pull request is now in conflict... :(
+      label:
+        add: [ "has conflicts" ]
+
+  - name: resolved conflicts
+    conditions:
+      - -conflict
+      - label="has conflicts"
+      - -draft # filter-out GH draft PRs
+      - -merged # not merged yet
+      - -closed
+    actions:
+      label:
+        remove: [ "has conflicts" ]
+
+  - name: update PR
+    conditions:
+      - -conflict
+      - -draft # filter-out GH draft PRs
+      - base=master  # apply only on master
+      - -title~=(?i)wip # skip all PR that title contains “WIP” (ignoring case)
+      - "#approved-reviews-by>=2" # number of review approvals
+    actions:
+      update: {}
+
+  - name: add core reviewer
+    conditions:
+      - -conflict # skip if conflict
+      - -draft # filter-out GH draft PRs
+      - "#approved-reviews-by<2" # number of review approvals
+      - "#review-requested<1" # number of requested reviews
+    actions:
+      request_reviews:
+        teams:
+          - "@PyTorchLightning/core-contributors"

.github/stale.yml

@@ -9,7 +9,7 @@ exemptLabels:
   - pinned
   - security
 # Label to use when marking an issue as stale
-staleLabel: wontfix
+staleLabel: "won't fix"
 # Comment to post when marking an issue as stale. Set to `false` to disable
 markComment: >
   This issue has been automatically marked as stale because it has not had

.gitignore

@@ -1,3 +1,4 @@
+.lock
 lightning_logs
 
 # Byte-compiled / optimized / DLL files

Makefile

@@ -0,0 +1,28 @@
+.PHONY: test clean docs
+
+# assume you have installed need packages
+export SPHINX_MOCK_REQUIREMENTS=1
+
+test: clean
+	pip install -r requirements.txt
+	pip install -r requirements/test.txt
+	# install APEX, see https://github.com/NVIDIA/apex#linux
+
+	# use this to run tests
+	python -m coverage run --source flash -m pytest flash tests -v --flake8
+	python -m coverage report
+
+docs: clean
+	pip install --quiet -r requirements/docs.txt
+	python -m sphinx -b html -W docs/source docs/build
+
+clean:
+	rm -rf _ckpt_*
+	rm -rf ./lightning_logs
+	# clean all temp runs
+	rm -rf $(shell find . -name "mlruns")
+	rm -rf .mypy_cache
+	rm -rf .pytest_cache
+	rm -rf ./docs/build
+	rm -rf ./docs/source/**/generated
+	rm -rf ./docs/source/api

README.md

@@ -39,26 +39,35 @@
 
 </div>
 
+---
+## News
+[Read our launch blogpost](https://pytorch-lightning.medium.com/introducing-lightning-flash-the-fastest-way-to-get-started-with-deep-learning-202f196b3b98)
+
 ---
 
 ## Installation
 
 Pip / conda
 
 ```bash
-pip install lightning-flash
+pip install lightning-flash -U
 ```
 
-Master
+Pip from source
+
 ```bash
-pip install git+https://github.com/PytorchLightning/lightning-flash.git@master --upgrade
+# with git
+pip install git+https://github.com/PytorchLightning/lightning-flash.git@master
+# OR from an archive
+pip install https://github.com/PyTorchLightning/lightning-flash/archive/master.zip
 ```
 
-Source
-
+From source using `setuptools`
 ``` bash
+# clone flash repository locally
 git clone https://github.com/PyTorchLightning/lightning-flash.git
 cd lightning-flash
+# install in editable mode
 pip install -e .
 ```
 
@@ -383,6 +392,9 @@ Join our [Slack](https://join.slack.com/t/pytorch-lightning/shared_invite/zt-f6b
 ## Community
 For help or questions, join our huge community on [Slack](https://join.slack.com/t/pytorch-lightning/shared_invite/zt-f6bl2l0l-JYMK3tbAgAmGRrlNr00f1A)!
 
+## Citations
+We’re excited to continue the strong legacy of opensource software and have been inspired over the years by Caffee, Theano, Keras, PyTorch, torchbearer, and fast.ai. When/if a paper is written about this, we’ll be happy to cite these frameworks and the corresponding authors.
+
 ## License
 Please observe the Apache 2.0 license that is listed in this repository. In addition
 the Lightning framework is Patent Pending.

docs/source/conf.py

@@ -147,3 +147,11 @@ def _package_list_from_file(pfile):
 MOCK_PACKAGES = [PACKAGE_MAPPING.get(pkg, pkg) for pkg in MOCK_PACKAGES]
 
 autodoc_mock_imports = MOCK_PACKAGES
+
+# only run doctests marked with a ".. doctest::" directive
+doctest_test_doctest_blocks = ''
+doctest_global_setup = """
+import torch
+import pytorch_lightning as pl
+import flash
+"""

docs/source/custom_task.rst

@@ -4,7 +4,7 @@ Tutorial: Creating a Custom Task
 In this tutorial we will go over the process of creating a custom task,
 along with a custom data module.
 
-.. code:: python
+.. testcode:: python
 
     import flash
 
@@ -21,7 +21,7 @@ Here we create a basic linear regression task by subclassing
 ``flash.Task``. For the majority of tasks, you will likely only need to
 override the ``__init__`` and ``forward`` methods.
 
-.. code:: python
+.. testcode::
 
     class LinearRegression(flash.Task):
         def __init__(self, num_inputs, learning_rate=0.001, metrics=None):
@@ -67,7 +67,7 @@ for the prediction of diabetes disease progression. We can create this
 ``DataModule`` below, wrapping the scikit-learn `Diabetes
 dataset <https://scikit-learn.org/stable/datasets/toy_dataset.html#diabetes-dataset>`__.
 
-.. code:: python
+.. testcode::
 
     class DiabetesPipeline(flash.core.data.TaskDataPipeline):
         def after_uncollate(self, samples):

docs/source/installation.md

@@ -5,13 +5,11 @@ Flash is tested on Python 3.6+, and PyTorch 1.6
 ## Install with pip/conda
 
 ```bash
-pip install lightning-flash
+pip install lightning-flash -U
 ```
 
 ## Install from source
 
 ``` bash
-git clone https://github.com/PyTorchLightning/flash.git
-cd flash
-pip install -e .
+pip install git+https://github.com/PyTorchLightning/lightning-flash.git
 ```

docs/source/quickstart.rst

@@ -45,7 +45,9 @@ organized PyTorch with the unecessary engineering details abstracted away.
 
 When you need more flexibility you can build your own tasks or simply use Lightning directly.
 
-.. note:: Read :doc:`here <reference/flash_to_pl>` to understand when to use Flash vs Lightning.
+.. tip::
+
+    Read :doc:`here <reference/flash_to_pl>` to understand when to use Flash vs Lightning.
 
 ----
 
@@ -56,7 +58,7 @@ You can install flash using pip or conda:
 
 .. code-block:: bash
 
-    pip install pytorch-lightning-flash -U -pre
+    pip install pytorch-lightning-flash -U
 
 ------
 

docs/source/reference/image_classification.rst

@@ -124,7 +124,9 @@ Changing the backbone
 *********************
 By default, we use a `ResNet-18 <https://arxiv.org/abs/1512.03385>`_ for image classification. You can change the model run by the task by passing in a different backbone.
 
-.. note:: When changing the backbone, make sure you pass in the same backbone to the Task and the Data object!
+.. note::
+
+    When changing the backbone, make sure you pass in the same backbone to the Task and the Data object!
 
 .. code-block:: python
 

docs/source/reference/image_embedder.rst

@@ -36,11 +36,11 @@ Or on a random image tensor
 
 .. code-block:: python
 
-  # 2. Perform inference on an image file
-  import torch
-  images = torch.rand(32, 3, 224, 224)
-  embeddings = model.predict(images)
-  print(predictions)
+    # 2. Perform inference on an image file
+    import torch
+    images = torch.rand(32, 3, 224, 224)
+    embeddings = model.predict(images)
+    print(predictions)
 
 For more advanced inference options, see :ref:`predictions`.
 

docs/source/reference/summarization.rst

@@ -171,7 +171,7 @@ API reference
 SummarizationTask
 -----------------
 
-.. autoclass:: flash.text.seq2seq.summarization.model.SummarizationTask
+.. autoclass:: flash.text.SummarizationTask
     :members:
     :exclude-members: forward
 
@@ -180,6 +180,6 @@ SummarizationTask
 SummarizationData
 -----------------
 
-.. autoclass:: flash.text.seq2seq.summarization.data.SummarizationData
+.. autoclass:: flash.text.SummarizationData
 
-.. automethod:: flash.text.seq2seq.summarization.data.SummarizationData.from_files
+.. automethod:: flash.text.SummarizationData.from_files

docs/source/reference/tabular_classification.rst

@@ -39,12 +39,10 @@ We can create :class:`~flash.tabular.TabularData` from csv files using the :func
 * **numerical_input**- a list of the names of columns that contain numerical continuous data (floats)
 * **target**- the name of the column we want to predict
 
-.. tip:: you can pass in val_size and test_size to partition your training data into a separate validation and test set like so:
-
 
 Next, we create the :class:`~flash.tabular.TabularClassifier` task, using the Data module we created.
 
-.. code-block::
+.. code-block:: python
 
     import flash
     from flash import download_data

docs/source/reference/translation.rst

@@ -5,7 +5,7 @@ Translation
 ###########
 
 ********
-The task
+The Task
 ********
 
 Translation is the task of translating text from a source language to another, such as English to Romanian.
@@ -17,7 +17,7 @@ This task is a subset of Sequence to Sequence tasks, which requires the model to
 Inference
 *********
 
-The :class:`~flash.text.TranslationTask` is already pre-trained on [WMT16 English/Romanian](https://www.statmt.org/wmt16/translation-task.html), a dataset of English to Romanian samples, based on the Europarl corpora.
+The :class:`~flash.text.TranslationTask` is already pre-trained on `WMT16 English/Romanian <https://www.statmt.org/wmt16/translation-task.html>`_, a dataset of English to Romanian samples, based on the `Europarl corpora <http://www.statmt.org/europarl/>`_.
 
 Use the :class:`~flash.text.TranslationTask` pretrained model for inference on any string sequence using :func:`~flash.text.TranslationTask.predict`:
 
@@ -78,7 +78,7 @@ Say you want to finetune to your own translation data. We use the English/Romani
 
 In the above the input/target columns represent the English and Romanian translation respectively.
 
-All we need is three lines of code to train our model!
+All we need is three lines of code to train our model! By default, we use a `mBART <https://github.com/pytorch/fairseq/tree/master/examples/mbart/>`_ backbone for translation which requires a GPU to train.
 
 .. code-block:: python
 
@@ -125,7 +125,7 @@ To run the example:
 *********************
 Changing the backbone
 *********************
-By default, we use `mBART <https://github.com/pytorch/fairseq/tree/master/examples/mbart/>`_ model for translation. You can change the model run by passing in the backbone parameter.
+You can change the model run by passing in the backbone parameter.
 
 .. note:: When changing the backbone, make sure you pass in the same backbone to the Task and the Data object! Since this is a Seq2Seq task, make sure you use a Seq2Seq model.
 
@@ -153,7 +153,7 @@ API reference
 TranslationTask
 ---------------
 
-.. autoclass:: flash.text.seq2seq.translation.model.TranslationTask
+.. autoclass:: flash.text.TranslationTask
     :members:
     :exclude-members: forward
 
@@ -162,6 +162,6 @@ TranslationTask
 TranslationData
 ---------------
 
-.. autoclass:: flash.text.seq2seq.translation.data.TranslationData
+.. autoclass:: flash.text.TranslationData
 
-.. automethod:: flash.text.seq2seq.translation.data.TranslationData.from_files
+.. automethod:: flash.text.TranslationData.from_files

flash/core/classification.py

@@ -14,7 +14,6 @@
 from typing import Any, Union
 
 import torch
-from transformers.modeling_outputs import SequenceClassifierOutput
 
 from flash.core.data import TaskDataPipeline
 from flash.core.model import Task

flash/core/data/datamodule.py

@@ -1,5 +1,18 @@
+# Copyright The PyTorch Lightning team.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
 import os
-import warnings
+import platform
 from typing import Any, Optional
 
 import pytorch_lightning as pl
@@ -51,11 +64,11 @@ def __init__(
         self.batch_size = batch_size
 
         # TODO: figure out best solution for setting num_workers
-        # if num_workers is None:
-        #    num_workers = os.cpu_count()
         if num_workers is None:
-            #    warnings.warn("Could not infer cpu count automatically, setting it to zero")
-            num_workers = 0
+            if platform.system() == "Darwin":
+                num_workers = 0
+            else:
+                num_workers = os.cpu_count()
         self.num_workers = num_workers
 
         self._data_pipeline = None