INDEX

Explanations

No Explanations Found

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

Redditor

-0.81

 clipping

-0.76

oi

-0.73

Tube

-0.71

arers

-0.71

ettings

-0.70

aceae

-0.67

 Seym

-0.66

 Tart

-0.65

oda

-0.65

POSITIVE LOGITS

imir

0.66

enged

0.63

 withstand

0.62

chron

0.62

utenberg

0.61

ief

0.61

cipled

0.61

ive

0.60

fast

0.60

ixt

0.60

Activations Density 0.000%

No Known Activations

This feature has no known activations.