facebookresearch / facebookresearch/detectron2
request for TEST.AMP.ENABLED
- Dominant language
- Python
- Stars
- 34.7k
- Forks
- 7.9k
- PR merge metrics
- No merged PRs in 30d
Description
## 🚀 Feature
A clear and concise description of the feature proposal.
## Motivation & Examples
Tell us why the feature is useful.
* Useful when accelerating model in inference time.
Describe what the feature would look like, if it is implemented.
Best demonstrated using **code examples** in addition to words.
```
# detectron2 > evaluation > evaluator.py > inference_on_dataset
# Not validated
if cfg.TEST.AMP.ENABLED:
with autocast():
loss_dict = self.model(data)
```
Contributor guide
Research direction
Start in detectron2/evaluation/evaluator.py at inference_on_dataset and read how evaluation invokes self.model(data). Then trace how cfg.TEST settings are defined and used. Done means the proposed TEST.AMP.ENABLED behavior is clearly specified, implemented consistently in the inference path, and demonstrated with code or tests.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- computer-vision, machine-learning
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 25/100