Future Roadmap Suggestions
RogerChern opened this issue · 13 comments
RogerChern commented
- Rewrite proposal generation from CUDA to Python
Thanks to apache/mxnet#14363, MXNet CustomOp now works parallelly. This enables us to migrate more CUDA codes to Python to ease the modification of core operators.
RogerChern commented
- Fix RoiAlignv2 compilation failure introduced by apache/mxnet#14270 in rogerchen/cuda9.0-cudnn7-py3:simpledet
RogerChern commented
- TVM deployment
RogerChern commented
- multi-node sync bn
RogerChern commented
- horovod distributed training
RogerChern commented
- speedup loader with apache/mxnet#14733
suman-19 commented
Rotation Invariant Object Detection to train on DOTA dataset.
RogerChern commented
- Support NDArray and Symbol dual modes for user-friendly debug.
RogerChern commented
- refactor code with apache/mxnet#14638
RogerChern commented
- refactor code with apache/mxnet#13985
RogerChern commented
- file corruption check for #182
RogerChern commented
- wrap the contrib RoIAlign for our use
Our one is written in quite an early day and is slightly different from the official Detectron implementation. Also, ours lacks FP16 support, sampling ratio, and PS variant.
RogerChern commented
pinning memory for loader
xchani commented
- model parallelism with pipeline