example_commands.txt

Some example commands which are tested fine on Stanford ICME:

We have 99 exodus files in the input directory: hdfs://icme-hadoop1.localdomain/user/yangyang/simform/

First, we setup database, and get the Makefile 'runs':

	make setup_database name=runs variable=TEMP dir=hdfs://icme-hadoop1.localdomain/user/yangyang/simform/

After we get 'runs', we can do the following things:

	make -f runs preprocess

Convert exodus files to sequence files:

	make -f runs convert timestepfile=timesteps.txt 
	(using normalized timesteps, time: 20min36s)
    
	make -f runs convert exodus2seq_output=hdfs://icme-hadoop1.localdomain/user/yangyang/simform/output/data.seq2/
	(without using normalized timesteps, time: 20min24s)

Compute the global variance:

    make -f runs var
    
Output the global variance:

    make -f runs outputvar

Predict without SVD:

	make -f runs predict design=design_points.txt points=new_points.txt
	(design points:99 new points:10, time:16min2s)

	make -f runs seq2exodus output_name=thermal_maze_interpolation
	(output 10 new interpolation exodus files, time: 4min45s)
    

Predict with SVD (we can get not only interpolation value but also estimated errors):

	Convert sequenc files to matrix sequence files:
	make -f runs seq2mseq
	(time:2hr...)

	Compute the SVD model:
	make -f runs model numExodusfiles=99
	(step1: full1, 44min46s
	 step2: full2  1min
	 step3: full3  4min6s
	 step4: TSMatMul 3min42s
	 total time: 54min
    
	Predict with SVD singular values and vectors (before this steps, we should create weights.txt first):
	make -f runs predict weights=weights.txt SVD=True    
	(new points:2, time: 22min26s)

	make -f runs seq2exodus output_name=thermal_maze_interpolation SVD=True
	(output 2 new interpolation exodus file, time: 2min40s)