INDEX

Explanations

No Explanations Found

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.hook_resid_pre

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.1.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

eed

-0.65

ettes

-0.63

arthed

-0.60

[_

-0.60

aze

-0.59

 Downtown

-0.57

val

-0.56

elta

-0.56

luc

-0.54

izon

-0.54

POSITIVE LOGITS

 Abbey

0.70

porting

0.70

Thom

0.70

udic

0.68

ERC

0.66

Magikarp

0.64

urden

0.63

oner

0.63

 Thomson

0.63

responsible

0.63

Activations Density 0.000%

No Known Activations

This feature has no known activations.