Poutyne

- A progress bar is now set on validation a model (similar to training). It is disableable by passing `progress_options=dict(show_on_valid=False)` in the `fit*` methods.
- A progress bar is now set testing a model (similar to training). It is disableable by passing `verbose=False` in the `evaluate*` methods.
- A new notification callback [`NotificationCallback`](https://poutyne.org/callbacks.html#poutyne.NotificationCallback) allowing to received message at specific time (start/end training/testing an at any given epoch).
- A new logging callback, [`MLflowLogger`](https://poutyne.org/callbacks.html#poutyne.MLFlowLogger), this callback allows you to log experimentation configuration and metrics during training, validation and testing.
- Fix bug where [`evaluate_generator`](https://poutyne.org/model.html#poutyne.Model.evaluate_generator) did not support generators with StopIteration exception.
- Experiment now has a [`train_data`](https://poutyne.org/experiment.html#poutyne.Experiment.train_data) and a [`test_data`](https://poutyne.org/experiment.html#poutyne.Experiment.test_data) method.
- The [Lambda layer](https://poutyne.org/layers.html#poutyne.Lambda) now supports multiple arguments in its forward method.

- A `device` argument is added to [`Model`](https://poutyne.org/model.html#poutyne.Model).
- The argument `optimizer` of [`Model`](https://poutyne.org/model.html#poutyne.Model) can now be a dictionary. This allows to pass different argument to the optimizer, e.g. `optimizer=dict(optim='sgd', lr=0.1)`.
- The progress bar now uses 20 characters instead of 25.
- The progress bar is now more fluid since partial blocks are used allowing increments of 1/8th of a block at once.
- The function [`torch_to_numpy`](https://poutyne.org/utils.html#poutyne.torch_to_numpy) now does .detach() before .cpu(). This might slightly improves performances in some cases.
- In Experiment, the [`load_checkpoint`](https://poutyne.org/experiment.html#poutyne.Experiment.load_checkpoint) method can now load arbitrary checkpoints by passing a filename instead of the usual argument.
- Experiment now has a `train_dataset` and a `test_dataset` method.
- Experiment is not considered a beta feature anymore.

**Breaking changes:**

- In [`evaluate`](https://poutyne.org/model.html#poutyne.Model.evaluate), `dataloader_kwargs` is now a dictionary keyword argument instead of arbitrary keyword arguments. Other methods are already this way. This was an oversight of the last release.

- There is now a batch metric [`TopKAccuracy`](https://poutyne.org/metrics.html#poutyne.TopKAccuracy) and it is possible to use them as strings for `k` in 1 to 10 and 20, 30, …, 100, e.g. `'top5'`.
- Add [`fit_dataset`](https://poutyne.org/model.html#poutyne.Model.fit_dataset) , [`evaluate_dataset`](https://poutyne.org/model.html#poutyne.Model.evaluate_dataset) and [`predict_dataset`](https://poutyne.org/model.html#poutyne.Model.predict_dataset) methods which allow to pass PyTorch Datasets and creates DataLoader internally. Here is [an example with MNIST](https://github.com/GRAAL-Research/poutyne/blob/master/examples/basic_mnist_classification.py) .
- Colors now work correctly in Colab.
- The default colorscheme was changed so that it looks good in Colab, notebooks and command line. The previous one was not readable in Colab.
- Checkpointing callbacks now don't use the Python [`tempfile` package](https://docs.python.org/3/library/tempfile.html) anymore for the temporary file. The use of this package caused problem when the temp filesystem was not on the same partition as the final destination of the checkpoint. The temporary file is now created at the same place as the final destination. Thus, in most use cases, this will render the use of the `temporary_filename` argument not necessary. The argument is still available for those who need it.
- In Experiment, it is not possible to call the method `test` when training without logging.

Update following bug in new PyTorch version: https://github.com/pytorch/pytorch/issues/47007

- Output is now very nicely colored and now has a progress bar. Both are disableable with the `progress_options` arguments. The `colorama` package needs to be installed to have the colors. See the documentation of the [fit](https://poutyne.org/model.html#poutyne.Model.fit) method for details.
- Multi-GPU support: Uses `torch.nn.parallel.data_parallel` under the hood.
- Huge update to the documentation with a documentation of metrics and a lot of examples.
- No need to import `framework` anymore. Everything now can be imported from `poutyne`directly, i.e. `from poutyne import whatever_you_want`.
- [`PeriodicSaveCallbacks`](https://poutyne.org/callbacks.html#poutyne.PeriodicSaveCallback) (such as [`ModelCheckpoint`](https://poutyne.org/callbacks.html#poutyne.ModelCheckpoint)) now has a flag `keep_only_last_best` which allow to only keep the last best checkpoint even when the names differ between epochs.
- [`FBeta`](https://poutyne.org/metrics.html#poutyne.FBeta) now supports an `ignore_index` as in `nn.CrossEntropyLoss`.
- Epoch metrics strings `'precision'` and `'recall'` now available directly without instantiating `FBeta`.
- Better ETA estimation in output by weighting more recent batches than older batches.
- Batch metrics [`acc`](https://poutyne.org/metrics.html#poutyne.acc) and [`bin_acc`](https://poutyne.org/metrics.html#poutyne.bin_acc) now have class counterparts [`Accuracy`](https://poutyne.org/metrics.html#poutyne.Accuracy) and [`BinaryAccuracy`](https://poutyne.org/metrics.html#poutyne.BinaryAccuracy) in addition to a `reduction` keyword argument as in PyTorch.
- Various bug fixes.

- Add new callback methods `on_test_*` to callbacks. Callback can now be passed to the `evaluate*` methods.
- New epoch metrics for scikit-learn functions ( See [documentation of SKLearnMetrics](https://poutyne.org/metrics.html#poutyne.framework.metrics.SKLearnMetrics)).
- It is now possible to return multiple metrics for a single batch metric function or epoch metric object. Furthermore, their names can be changed. (See note in [documentation of Model class](https://poutyne.org/model.html#poutyne.framework.Model))
- Computation of batch size is now added for dictionnary inputs and outputs. ( See [documentation of the new method `get_batch_size`](https://poutyne.org/model.html#poutyne.framework.Model.get_batch_size))
- Add a lot of type hinting.

**Breaking changes:**

- Ground truths and predictions returned by `evaluate_generator` and `predict_generator` are going to be concatenated except when inside custom objects in the next version. A warning is issued in those methods. If the warning is disabled as instructed, the new behavior takes place. (See documentation of [evaluate_generator](https://poutyne.org/model.html#poutyne.framework.Model.evaluate_generator) and [predict_generator](https://poutyne.org/model.html#poutyne.framework.Model.predict_generator))
- Names of methods `on_batch_begin` and `on_batch_end` changed to `on_train_batch_begin` and `on_train_batch_end` respectively. When the old names are used, a warning is issued with backward compatibility added. This backward compatibility will be removed in the next version.
- `EpochMetric` classes now have an obligatory reset method.
- Support of Python 3.5 is dropped. (Anyway, PyTorch was already not supporting it either)