ignite

ignite copied to clipboard

Update for `timeline_and_event` image

1

### 📚 Documentation When I first read about [timeline and events in PyTorch-Ignite concepts](https://pytorch.org/ignite/master/concepts.html#timeline-and-events), I noticed I had to rotate my head around 30 degree to read [`timeline_and_event` image](https://pytorch.org/ignite/master/_static/img/concepts/timeline_and_events.png). Should...

ydcjeff

Improve idist for gather using nccl and reduce for gloo + gpu

1

## 🚀 Feature Consider the following piece of code ```python def write_preds_to_file(predictions, filename): prediction_tensor = torch.tensor(predictions) prediction_tensor = idist.all_gather(prediction_tensor) if idist.get_rank() == 0: torch.save(prediction_tensor, filename) ``` The `idist.all_gather()` is used...

sdesrozis

[docs] only build and deploy master

4

## 📚 Documentation We are currently building the docs from all tags / releases using master template and configs. And there are a bunch of warnings for past releases. This...

ydcjeff

Mean Average Precision

3

Fixes #{520} Description: Created a COCO Style Implementation for mAP metric. Both of them are calculated at once since both require similar calculation and have similar use case. The overall...

gucifer

Update class.rst

5

Fixes #{issue number} Description: Check list: - [ ] New tests are added (if a new feature is added) - [ ] New doc strings: description and/or example code are...

ydcjeff

Addresses #336 #2007 Description: Prevents metric from being computed before updated using `__new__` magic. `__new__` is called by default on instance creation before `__init__` call. Thus, we wrap methods before...

trsvchn

Issue 1986 Checkpointing write & remove order change

1

Fixes #1986 Description: Former "remove & write" implementation order is reversed to "write & remove", api call has been added for ClearMLSaver remove functionality. Check list: - [ ] New...

devrimcavusoglu

Checkpoint stateful handlers and metrics

3

## 🚀 Feature Things that are attached to the `Engine` might have state that would ideally be checkpointed and restored using as part of the `Engine`'s `state_dict`. An example is...

amatsukawa

Add device arg to Engine

3

## 🚀 Feature The idea is to simplify `processing_function` with an optional `device` arg to `Engine`: ```diff def train_step(engine, batch): x, y = batch[0], batch[1] - if x.device != device:...

vfdev-5

Wrong docstrings for Metric `update` method's input type

3

## 📚 Documentation For example, [accuracy documentation](https://pytorch.org/ignite/metrics.html#ignite.metrics.Accuracy) says that > update must receive output of the form `(y_pred, y)` or `{'y_pred': y_pred, 'y': y}`. which is incorrect as it can...

vfdev-5