Megvii-BaseDetection / Megvii-BaseDetection/YOLOX
High memory consumption with all YOLOX variants
Nobody has claimed this yet.
- Dominant language
- Python
- Stars
- 10.6k
- Forks
- 2.5k
- PR merge metrics
- No merged PRs in 30d
Description
The memory consumption seems extremely high during training. For all the variants it is occupying almost entire capacity of GPU, I'm training with batch size of 16 on a 32G GPU. Ultralytics models had a comparatively low memory consumption around 15G for X variant, and 12G for L variant. Is this behavior normal, am I missing something.
My training params for reference
batch_size=16
num_workers=4
image size = 640,640
cache is turned off
Contributor guide
No contributing guide indexed for this repository
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Use the reported training setup—batch size 16, four workers, 640x640 images, and cache off—to reproduce GPU memory usage across YOLOX variants. Compare the results with the report and investigate the training path for the source of the allocation. Done means establishing whether the usage is expected or producing a verified fix or narrowed reproduction.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python, pytorch
- Domain
- machine-learning, performance
- Issue type
- Bug
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 25/100