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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

hani

-0.86

fters

-0.78

smanship

-0.73

Store

-0.68

isbury

-0.68

aptic

-0.65

ffe

-0.64

\\\\\\\\

-0.64

Mi

-0.62

brance

-0.62

POSITIVE LOGITS

 Wolverine

0.81

emouth

0.80

 Hawaiian

0.76

 Brach

0.71

 SEAL

0.67

ora

0.65

 Zhou

0.65

uyomi

0.65

NYU

0.65

Rug

0.64

Activations Density 0.000%

No Known Activations

This feature has no known activations.