Can you share the training statements for other phases?
Opened this issue · 6 comments
In the command you share publicly, you can get to the training statement for Phase 2，Thank you
Command for other stages can't be found, can you share that?
Sure, you could simply change the config name, like:
CUDA_VISIBLE_DEVICES=0,1,2,3 PORT=29500 bash tools/dist_train.sh configs/centernet_AERIS/centernet_res18_stage1.py 4
Config names find in here: https://github.com/cuiziteng/ECCV_AERIS/tree/master/configs
Sure, you could simply change the config name, like:当然，您可以简单地更改配置名称，例如：
CUDA_VISIBLE_DEVICES=0,1,2,3 PORT=29500 bash tools/dist_train.sh configs/centernet_AERIS/centernet_res18_stage1.py 4CUDA_VISIBLE_DEVICES=0,1,2,3 端口=29500 bash 工具/dist_train.sh配置/centernet_AERIS/centernet_res18_stage1.py 4
Config names find in here: https://github.com/cuiziteng/ECCV_AERIS/tree/master/configs配置名称可在此处找到： https://github.com/cuiziteng/ECCV_AERIS/tree/master/configs
训练好的模型会保存在哪个文件夹呢？如何修改保存路径呢？
您可以自己设置保存路径的位置，比如在训练指令后加上--work-dir "Your_path"，当然mmdetection也会自己默认给你一个保存地址
例如：
CUDA_VISIBLE_DEVICES=0,1,2,3 PORT=29500 bash tools/dist_train.sh configs/centernet_AERIS/centernet_res18_stage1.py --work-dir work_dirs
这样就会保存在你当前目录下的work_dirs文件夹
您可以自己设置保存路径的位置，比如在训练指令后加上--work-dir "Your_path"，当然mmdetection也会自己默认给你一个保存地址
例如： CUDA_VISIBLE_DEVICES=0,1,2,3 PORT=29500 bash tools/dist_train.sh configs/centernet_AERIS/centernet_res18_stage1.py --work-dir work_dirsCUDA_VISIBLE_DEVICES=0,1,2,3 端口=29500 bash tools/dist_train.sh configs/centernet_AERIS/centernet_res18_stage1.py --work-dir work_dirs
这样就会保存在你当前目录下的work_dirs文件夹
您好，tools目录下有一个的train_SR.py，超分模型是需要单独训练出来吗？还是直接训练centernet_res18_stage1.py 就行呢？
您好，按照指示来就好，train_SR.py这个是用来调整不同模块学习率的，
ECCV_AERIS/mmdet/apis/train.py
Line 211 in 42ab63c