train=trainName, syr=2015, eyr=2015, var='varLst_Forcing', varC='varConstLst_Noah', dr=0.6, modelOpt='relu', model='cudnn', loss='sigma') k = 0 for j, i in zip(C1Lst, C2Lst): opt['out'] = trainName + \ '_y15_Forcing_dr60_invGamma_'+str(j)+'_'+str(i) opt['lossPrior'] = 'invGamma+' + str(j) + '+' + str(i) runTrainLSTM.runCmdLine(opt=opt, cudaID=k % 3, screenName=opt['lossPrior']) # rnnSMAP.funLSTM.trainLSTM(opt) k = k + 1 ################################################# if 'test' in doOpt: torch.cuda.empty_cache() dsLst = list() statErrLst = list() statSigmaLst = list() statConfLst = list() statNormLst = list() for k in range(0, nCase): out = outLst[k] ds = rnnSMAP.classDB.DatasetPost(rootDB=rootDB,
rootOut=rootOut, syr=2015, eyr=2015, var='varLst_Forcing', varC='varConstLst_Noah', train='CONUSv4f1', dr=0.5, modelOpt='relu', model='cudnn', loss='sigma', ) for k in range(0, len(drLst)): opt['dr'] = drLst[k] opt['out'] = 'CONUSv4f1_y15_Forcing_dr' + drStrLst[k] cudaID = k % 3 runTrainLSTM.runCmdLine(opt=opt, cudaID=cudaID, screenName=opt['out']) ################################################# if 'test' in doOpt: predField = 'LSTM' targetField = 'SMAP' dsLst = list() statErrLst = list() statSigmaLst = list() statConfLst = list() for k in range(0, len(drStrLst)): if drStrLst[k] is '50': out = 'CONUSv4f1_y15_Forcing' else: out = 'CONUSv4f1_y15_Forcing_dr' + drStrLst[k] testName = testName
rootOut=rnnSMAP.kPath['OutSigma_L3_NA'], syr=2015, eyr=2015, var='varLst_soilM', varC='varConstLst_Noah', dr=0.5, modelOpt='relu', model='cudnn', loss='sigma') cudaIdLst = [1, 2] for k in range(0, len(hucLst)): trainName = hucLst[k] + '_v2f1' opt['train'] = trainName opt['out'] = trainName + '_y15_soilM' print(trainName) runTrainLSTM.runCmdLine(opt=opt, cudaID=cudaIdLst[k], screenName=trainName) ################################################# # Test rootOut = rnnSMAP.kPath['OutSigma_L3_NA'] rootDB = rnnSMAP.kPath['DB_L3_NA'] dsTuple = ([], [], []) dsTuple2 = ([], []) for k in range(0, len(hucLst)): trainName = hucLst[k] + '_v2f1' out = trainName + '_y15_soilM' rootOut = rnnSMAP.kPath['OutSigma_L3_NA'] outLst = [ trainName + '_y15_soilM', trainName + '_y15_soilM', 'CONUSv2f1_y15_soilM'