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

 mathemat

-0.82

 retina

-0.72

 Canaver

-0.70

XV

-0.67

ãĥĩãĤ£

-0.63

_(

-0.63

illus

-0.63

 forestry

-0.62

XIV

-0.61

bip

-0.61

POSITIVE LOGITS

hetti

0.74

--------------------------------------------------------

0.69

ushing

0.66

assetsadobe

0.66

aea

0.66

baugh

0.65

xual

0.63

bol

0.62

ploy

0.61

hid

0.59

Activations Density 0.000%

No Known Activations

This feature has no known activations.