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

 tast

-0.73

aign

-0.67

 muff

-0.64

ream

-0.64

cow

-0.63

ħĭ

-0.63

 punch

-0.63

nn

-0.62

 transport

-0.62

cellent

-0.62

POSITIVE LOGITS

bral

0.74

PsyNetMessage

0.72

enegger

0.69

 Commodore

0.68

 Forbes

0.66

Commerce

0.66

CSI

0.63

 Able

0.63

ogene

0.63

 Astro

0.61

Activations Density 0.000%

No Known Activations

This feature has no known activations.