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

Champ

-0.69

dayName

-0.66

ATES

-0.66

doors

-0.65

 Prism

-0.64

 coli

-0.63

opolis

-0.62

hands

-0.61

Ross

-0.61

 precincts

-0.60

POSITIVE LOGITS

 precedence

0.66

ime

0.66

lite

0.66

ayn

0.65

ize

0.64

endor

0.64

endez

0.64

 clipboard

0.63

outed

0.61

ICLE

0.61

Activations Density 0.000%

No Known Activations

This feature has no known activations.