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

 architects

-0.72

pelled

-0.72

olding

-0.70

BILITIES

-0.70

anium

-0.67

unity

-0.65

lu

-0.65

unct

-0.63

 anim

-0.63

blem

-0.62

POSITIVE LOGITS

 helicop

0.88

 looph

0.75

 psychiat

0.75

ĺħ

0.75

 Posts

0.74

manship

0.70

ħĭ

0.70

ÃĥÃĤ

0.66

erva

0.66

ŃĶ

0.66

Activations Density 0.000%

No Known Activations

This feature has no known activations.