Eunwoo Shin
Eunwoo Shin
This PR fixes a bug which is occurred when HPO runs without batch size in search space.
### Summary This PR registers signal handler for mp.Manager in MemCacheHandlerForMP. If signal handler doesn't exist, manager process remains alive even if main process is killed by signal. It applied...
**Describe the bug** When training tilling ins-seg model with "coco decremental dataset" which is used for regression test dataset, an error is raised due to index out of range error...
**Describe the bug** GPU memory increases although input image size decreases in instance segmentation. When I trained both instance segmentation model with 700x700 input size using wgisd-5 dataset, GPU memory...
### Summary This PR adds new HPO feature. New feature can adapt max value of HPO batch size search space automatically. It provides user an easy way to avoid OOM...
### Summary Currently, OTX uses own model_api model wrapper for action classification task. It's implemented at model_api 0.2.1, so make OTX use it and remove unnecessary code. ### Integration test...
### Summary This PR inlcudes * change `update_mem_cache_img_max_size` to `update_mem_cache_img_min_size` * set `update_mem_cache_img_min_size` to value of `input_size` if it's `None`. * set `update_mem_cache_img_min_size` to None if tiling ### Details ####...
### Summary ### How to test ### Checklist - [ ] I have added unit tests to cover my changes. - [ ] I have added integration tests to cover...