lstm1024H0P2L_emb1024_adam_lr0.001_bs20_bptt30_tie_residual_glu_1tokens
% PPL
valid (baseline): 87.99
valid (cache size: 100): 79.58
valid (cache size: 500): 77.36
test (baseline): 86.06
test (cache size: 100): 79.12
test (cache size: 500): 76.94
