X-Git-Url: https://fleuret.org/cgi-bin/gitweb/gitweb.cgi?a=blobdiff_plain;f=README.txt;h=a4cd46b3ab7feaf839fac3b94e8268eb2d9f01ca;hb=291c38d093894d46fba6eb45f82e5b65a2a1cb8b;hp=231d292110ec4d88e7a22e2273683a8f7c53e0bb;hpb=7c8049b2dde68455147c74c6245fc5d44aa03077;p=picoclvr.git diff --git a/README.txt b/README.txt index 231d292..a4cd46b 100644 --- a/README.txt +++ b/README.txt @@ -1,10 +1,12 @@ +====================================================================== For the stack experiment: ./main.py --task=stack Takes ~1h10min on a 4090. +====================================================================== For the arithmetic expressions experiments # 38M parameters / 250k samples @@ -14,4 +16,5 @@ For the arithmetic expressions experiments # 352M parameters / 2.5M samples, reaches 99.80% after 12 epochs, the learning rate schedule is obviously terrible -./main.py --task=expr --nb_blocks=48 --result_dir=results_expr_48b_d1024_2.5M --dim_model=1024 --nb_train_samples=2500000 +./main.py --task=expr --nb_blocks=48 --dim_model=1024 --nb_train_samples=2500000 --result_dir=results_expr_48b_d1024_2.5M +======================================================================