INDEX

Explanations

No Explanations Found

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

Īè

-0.64

auntlets

-0.63

 amygdala

-0.62

tip

-0.61

inker

-0.61

plates

-0.59

etsy

-0.59

DN

-0.59

iman

-0.59

NN

-0.59

POSITIVE LOGITS

 serious

1.23

serious

1.01

 Serious

0.82

 seriousness

0.80

uncture

0.76

ogging

0.69

 Cyan

0.68

alian

0.66

 Fantasy

0.66

arnaev

0.65

Activations Density 0.000%

No Known Activations

This feature has no known activations.