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

advertisement

-0.68

rences

-0.65

 electors

-0.64

 Garrison

-0.63

secution

-0.62

patch

-0.62

 annotations

-0.62

 deposition

-0.61

 multiplication

-0.61

 evaluations

-0.61

POSITIVE LOGITS

imar

0.86

ESCO

0.80

imately

0.73

iously

0.72

LEY

0.70

bourg

0.69

wine

0.67

idate

0.67

isconsin

0.67

berry

0.65

Activations Density 0.000%

No Known Activations

This feature has no known activations.