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

oft

-0.73

 Beckham

-0.70

lain

-0.68

 Jelly

-0.67

ters

-0.67

Champ

-0.65

ijn

-0.64

oqu

-0.63

Om

-0.63

 Guardiola

-0.63

POSITIVE LOGITS

WARE

0.87

hol

0.83

 neighbors

0.77

 neighbor

0.76

 IMAGES

0.76

 Azerb

0.76

 arrang

0.75

 mathemat

0.75

 opio

0.74

everal

0.73

Activations Density 0.000%

No Known Activations

This feature has no known activations.