INDEX

Explanations

words related to programming or coding concepts

New Auto-Interp

Top Features by Cosine Similarity

Configuration

ckkissane/attn-saes-gpt2-small-all-layers/gpt2-small_L5_Hcat_z_lr1.20e-03_l11.00e+00_ds49152_bs4096_dc1.00e-06_rsanthropic_rie25000_nr4_v9.pt

Prompts (Dashboard)

36,864 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

49,152

Data Type

float32

Hook Name

blocks.5.attn.hook_z

Hook Layer

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Activation Function

relu

Embeds

IFrame

Link

Head Attr Weights

0:0.02

1:0.02

2:0.10

3:0.07

4:0.06

5:0.05

6:0.05

7:0.08

8:0.04

9:0.07

10:0.31

11:0.08

Negative Logits

 senses

-2.45

 absorb

-2.25

 fluent

-2.18

 intuitive

-2.15

 bonded

-2.15

 fingertips

-2.02

 handy

-2.01

eger

-2.01

 grades

-2.00

 sensing

-2.00

POSITIVE LOGITS

[|

2.30

cause

2.30

onies

2.23

Christmas

2.22

 Anthem

2.18

soDeliveryDate

2.16

fell

2.02

 Title

2.02

Theme

1.99

cens

1.96

Activations Density 0.000%

words related to programming or coding concepts

No Known Activations