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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

INESS

-0.79

etheless

-0.70

 Chero

-0.68

INGTON

-0.66

xon

-0.66

åĽ

-0.65

ä

-0.64

yout

-0.64

ingly

-0.63

soDeliveryDate

-0.63

POSITIVE LOGITS

 metabol

0.72

 menus

0.65

 bacterial

0.64

Â·

0.63

Sic

0.61

 immigrant

0.61

 breakfast

0.60

enery

0.59

 infections

0.59

Ion

0.59

Activations Density 0.000%

No Known Activations

This feature has no known activations.