|
|
a3bad799e3
|
sae werks with deep loss
|
2023-12-23 15:32:18 +01:00 |
|
|
|
ecd7d3bd65
|
tensorboard, low l1, lower lr
|
2023-12-14 13:44:35 +01:00 |
|
|
|
bc7647cb43
|
big search for good learning rate
|
2023-12-13 20:42:30 +01:00 |
|
|
|
63048b1915
|
gradually getting there, have yet to get a single trained sae
|
2023-10-19 17:45:43 +02:00 |
|
