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

 weap

-0.74

 mosqu

-0.67

 istg

-0.66

 cannabin

-0.65

âĹ¼

-0.63

hold

-0.63

—-

-0.60

 vain

-0.60

 Allaah

-0.60

 horm

-0.60

POSITIVE LOGITS

1.43

lio

0.83

uity

0.83

uay

0.79

ued

0.79

uum

0.79

hess

0.78

uable

0.76

ullivan

0.75

uve

0.75

Activations Density 0.000%

No Known Activations

This feature has no known activations.