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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 Tsukuyomi

-0.67

pired

-0.65

 Tens

-0.65

 Quartz

-0.65

destruct

-0.64

achus

-0.63

raph

-0.62

 mosqu

-0.61

Roz

-0.60

tions

-0.60

POSITIVE LOGITS

armac

0.81

Done

0.73

ionic

0.69

oyal

0.69

haw

0.68

Roll

0.68

AIR

0.68

UGE

0.66

arde

0.65

ighty

0.65

Activations Density 0.000%

No Known Activations

This feature has no known activations.