Tom Lieberum
9B: Add sparsity lambdas for residual stream and clean up the feature splitting suite so that we only have SAEs with learning rate 7e-5
745af8a
- embedding
- layer_0
- layer_1
- layer_10
- layer_11
- layer_12
- layer_13
- layer_14
- layer_15
- layer_16
- layer_17
- layer_18
- layer_19
- layer_2
- layer_20
- layer_21
- layer_22
- layer_23
- layer_24
- layer_25
- layer_26
- layer_27
- layer_28
- layer_29
- layer_3
- layer_30
- layer_31
- layer_32
- layer_33
- layer_34
- layer_35
- layer_36
- layer_37
- layer_38
- layer_39
- layer_4
- layer_40
- layer_41
- layer_5
- layer_6
- layer_7
- layer_8
- layer_9
-
1.84 kB
-
920 Bytes
-
1.36 kB