X-Git-Url: https://fleuret.org/cgi-bin/gitweb/gitweb.cgi?a=blobdiff_plain;f=README.txt;fp=README.txt;h=bf6180a1c5b3edbacc5da38493528b921bdcd871;hb=45f841397dd6f8fb163f2c5b0793cdc250072c73;hp=223565e63ad6236d33f90fa38e18d8b15e9d3193;hpb=703baba8b6619c57562e1d622be75c8b409659e5;p=picoclvr.git diff --git a/README.txt b/README.txt index 223565e..bf6180a 100644 --- a/README.txt +++ b/README.txt @@ -1,5 +1,16 @@ + For the stack experiment: -./main.py --task=stack --stack_fraction_values_for_train=0.75 --stack_nb_stacks=3 +./main.py --task=stack Takes ~1h10min on a 4090. + +For the arithmetic expressions experiments + +# 38M parameters / 250k samples + +./main.py --task=expr + +# 352M parameters / 2.5M samples + +./main.py --task=expr --nb_blocks=48 --result_dir=results_expr_48b_d1024_2.5M --dim_model=1024 --nb_train_samples=2500000 --learning_rate_schedule="1:2e-5,3:4e-6"