srush / annotated-s4

Implementation of https://srush.github.io/annotated-s4
https://srush.github.io/annotated-s4
MIT License
468 stars 60 forks source link

About setting HYDRA_FULL_ERROR=1 #79

Open muxiddin19 opened 5 months ago

muxiddin19 commented 5 months ago

Nice work! I am facing trouble with running: python -m s4.train dataset=mnist layer=s4 train.epochs=100 train.bsz=128 model.d_model=128 model.layer.N=64

It gives: -3.1771252, -6.8962207, -4.7297306, -3.454745 , -5.095578 , -5.370283 , -4.856248 , -4.50486 , -4.5061316, -6.3002734, -3.7687006, -5.7974706, -3.859034 , -2.3242874, -5.817657 , -5.110781 , -4.725958 , -6.5070057, -6.8027916, -3.445009 , -2.7452703, -5.913268 ]], dtype=float32), }, }, }).

Set the environment variable HYDRA_FULL_ERROR=1 for a complete stack trace.

However, I have not fixed it yet. Can you please guide me?