The genTen program generates sparse tensors with several given features. It adopts a combination of normal distribution and log-normal distribution.
To compile the program, use the command
./compile.sh
USAGE:
./genten order sizes[] [options]
-d density : nonzero ratio [default: 0.001]
-f density_fiber : nonzero fiber ratio for mode-(M) fibers [default: 0.01]
-s density_slice : nonzero slice ratio for mode-(M-1,M) slices [default: 1.0]
-c cv_fib_per_slc : coefficient of variation of fiber per slice values for mode-(M) fibers and mode-(M-1,M) slices [default: 0.5]
-v cv_nz_per_fib : coefficient of variation of nonzero per fiber values for mode-(M) fibers [default: 0.5]
-i imbal_fib_per_slc : imbalance of fiber per slice values for mode-(M) fibers and mode-(M-1,M) slices. ( imbalance = max/avg-1 ) [default: max = size]
-b imbal_nz_per_fib : imbalance of nonzero per fiber values for mode-(M) fibers. ( imbalance = max/avg-1 ) [default: max = size]
-r random_seed : seed for randomness [default: 1]
-h print_header : whether print the header names for explaining the output values (1:yes, 0:no[default])
-p print_debug : whether print details at some steps for debugging (1:yes, 0:no[default])
-w write_tensor : whether write the generated tensor into a file or not (1:yes[default], 0:no)
-o outfile : the name of the output file to print out the generated tensor
Example run commands will be like the following:
./genten 3 100 100 100 -d 0.01 -f 0.1 -c 0.5 -v 0.5 -o ../sample_data/generated_100_3D.tns
./genten 4 100 100 100 100 -d 0.0001 -f 0.1 -c 0.5 -v 0.5 -o ../sample_data/generated_100_4D.tns
./genten 5 10 20 30 40 50 -d 0.0001 -f 0.001 -s 0.01 -c 0.5 -v 0.5 -o ../sample_data/generated_5d.tns
./genten 3 10000 15000 20000 -d 2E-09 -f 0.0002 -v 23.1 -s 0.8 -c 13.0 -o ../sample_data/sample_generated.tns
The output of the program is in the form (which can be printed by typing "-h 1" option) :
name seed order dim_0 dim_1 dim_2 slc_type nz_slc_cnt requested result density_slc requested result ratio distr_type no max_fib_per_slc requested result ratio imbal_fib_per_slc requested result ratio std_fib_per_slc requested result ratio cv_fib_per_slc requested result ratio avg_fib_per_slc requested result ratio nz_fib_cnt requested result density_fib requested result ratio_prev ratio_last distr_type no max_nz_per_fib requested result ratio imbal_nz_per_fib requested result ratio std_nz_per_fib requested result ratio cv_nz_per_fib requested result ratio avg_nz_per_fib requested result ratio nnz requested result density requested result ratio_prev ratio_last threads TIME time_slc time_fib time_nz time_nz_ind time_write time_total
generate_real_tensor_siblings.sh contains a list of commands to generate tensors that are similar to real tensors (in FROSTT and HaTeN2 datasets).
generate_example.sh is an extended version of these commands, which allows to generate tensors with different seeds.
The genten_naive_rand program generates simple naive random tensors with uniform distribution, this is just for comparison.