INDEX

Explanations

No Explanations Found

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

 payroll

-0.74

OFF

-0.72

senal

-0.72

 Remastered

-0.65

ierrez

-0.60

 scheme

-0.60

gap

-0.60

 recru

-0.59

 novelist

-0.59

 somew

-0.59

POSITIVE LOGITS

rium

0.90

imental

0.73

 Caesar

0.69

assetsadobe

0.68

esy

0.66

 flare

0.66

ILE

0.65

ffic

0.63

 caution

0.63

 frying

0.62

Activations Density 0.000%

No Known Activations

This feature has no known activations.