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

wcsstore

-0.67

Rin

-0.67

ETHOD

-0.63

chen

-0.63

Sek

-0.62

 Aqua

-0.61

âĸ¬âĸ¬

-0.61

================================================================

-0.60

SEA

-0.59

ealous

-0.58

POSITIVE LOGITS

 martyr

0.75

ioned

0.70

 polio

0.70

 veter

0.69

orthy

0.67

 Karachi

0.63

riots

0.63

obb

0.62

obal

0.62

Jews

0.62

Activations Density 0.000%

No Known Activations

This feature has no known activations.