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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 diapers

-0.71

 commercials

-0.68

iatrics

-0.67

soDeliveryDate

-0.67

thumbnails

-0.65

iverpool

-0.65

EED

-0.63

baby

-0.60

 yours

-0.60

 theirs

-0.60

POSITIVE LOGITS

tein

0.92

imet

0.71

Kut

0.68

 Forge

0.67

Ree

0.66

ene

0.66

ixt

0.66

etry

0.65

 Turing

0.63

cius

0.63

Activations Density 0.000%

No Known Activations

This feature has no known activations.