INDEX

Explanations

No Explanations Found

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.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

 Null

-0.69

 Troll

-0.69

hammer

-0.68

gate

-0.66

 Borders

-0.66

Reviewer

-0.66

 Aether

-0.65

 Krypt

-0.65

 Sunder

-0.64

 Dare

-0.64

POSITIVE LOGITS

 advert

0.73

 ejac

0.73

ĨĴ

0.70

PDATE

0.68

 lapt

0.67

ENE

0.66

 anecd

0.66

 hourly

0.66

iasm

0.63

 physiological

0.63

Activations Density 0.000%

No Known Activations

This feature has no known activations.