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

Bed

-0.76

Shape

-0.75

aez

-0.73

IES

-0.73

NetMessage

-0.70

>]

-0.70

Materials

-0.69

rea

-0.69

aria

-0.68

Desk

-0.66

POSITIVE LOGITS

 dilig

0.73

 Vaugh

0.70

 synopsis

0.67

 Constantin

0.63

 indicative

0.62

 Siber

0.61

 samurai

0.61

geist

0.60

 kicker

0.60

 popul

0.58

Activations Density 0.000%

No Known Activations

This feature has no known activations.