Check "unetr" and "resunet_se" models

[danifranco]

Seems that in these models some of the layers are not being updated (you can see it only in DDP setting). I'm talking about this line:

BiaPy/biapy/engine/base_workflow.py

Line 706 in 9efbbec

if self.cfg.MODEL.ARCHITECTURE.lower() in ["unetr", "resunet_se"]:

We should check which layers are those by using TORCH_DISTRIBUTED_DEBUG. For that we can call BiaPy like this:

TORCH_DISTRIBUTED_DEBUG=DETAIL python -u -m torch.distributed.run --nproc_per_node=3 --master-port=$(shuf -i 5000-10000 -n 1)  main.py                --config $job_cfg_file                --result_dir $result_dir                 --name $job_name                   --run_id $job_counter                 --gpu "1,2,3"

[danifranco]

Fixed in fd648f8 and 080d8f3 !