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

odox

-0.84

 disadvantages

-0.77

ties

-0.75

antine

-0.69

avier

-0.67

eer

-0.67

 pros

-0.65

usable

-0.65

 casualties

-0.64

 homework

-0.62

POSITIVE LOGITS

uity

0.76

apeake

0.68

 Peng

0.66

 stride

0.65

 Prescott

0.65

 Shutterstock

0.65

 Pengu

0.64

SHIP

0.64

law

0.63

doms

0.62

Activations Density 0.000%

No Known Activations

This feature has no known activations.