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

Negative Logits

workshop

-0.77

uddle

-0.68

SEA

-0.68

enture

-0.68

 concussion

-0.65

©¶æ

-0.64

 unnecess

-0.63

ulhu

-0.62

[|

-0.62

ĻĤ

-0.62

POSITIVE LOGITS

 Corinth

0.66

 stabil

0.65

 Evangel

0.64

opol

0.63

 Kens

0.63

 SOLD

0.63

 tails

0.63

 Ware

0.63

hon

0.62

Moy

0.62

Activations Density 0.000%

No Known Activations

This feature has no known activations.