Readme.paths

INPUT data :
/global/cscratch1/sd/balewski/srgan_cosmo2d_data
SMALL 512-cube data set
h5ls univL9cola_dm_202204_c30.h5
meta.JSON                Dataset {1}
test.hr                  Dataset {4608,  512, 512, 2}
val.hr                   Dataset {4608,  512, 512, 2}
train.hr                 Dataset {36864, 512, 512, 2}
The last dimR 0: zRed=50, 1:zRed=0.

Dataloader will compute lr-images in fly during loading


Output dir structure

***Every experiment,prediction writes all outputs into:
  $expPath/$expName/
   sub-dirs:  checkpoints/ snapshots/ tb_logs/
   files: sum_train.yaml pred-test-best521.h5

***Configuring train interactively***
train_dist.py --basePath [none/absol] --expName [exp12/jobId]
action:  expPath=basePath/expName
         create if not existing:  checkpoints/ snapshots/

***Configuring train Slumr job***
shell action:
    mapping: expName=$jobId , basePath=PickMe
    expPath=$basePath/$expName
    create $basePath if not existing
    copy code to $expPath
    cd $expPath
    train_dist.py --basePath $basePath --expName $jobId
     (the rest should work)

***Configuring predict ***
./predict.py --basePath $basePath --expName $expName --genSol $solName --outPath same
action:
     expPath=$basePath/$expName
     modelPath=$expPath/checkpoints/genSol [+.pth]
     outPath=expPath if 'same'
     read sum_train.yaml
     device=CPU
     create G-model from .py
     prime model state from $modelPath
     prime dataloader from trainMD from .yaml
     predict (nSamp)
     write h5 to $outPath

***Configuring analysis of predictions ***
./ana_sr2d.py  --dataPath [expPath]  --genSol $solNamee  --outPath out/
Alternative:
./ana_sr2d.py --expName exp04 --genSol last  
    actoion:  dataPath=dataPath+expName

./ana_power1.py --expName exp04 --genSol last