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

UES

-0.72

UE

-0.71

uing

-0.69

Tex

-0.67

Brazil

-0.66

 instance

-0.66

 ranc

-0.65

 Vald

-0.64

 Brazilian

-0.63

geist

-0.62

POSITIVE LOGITS

EStreamFrame

1.14

yrinth

1.03

anus

1.01

chwitz

0.77

 unfocusedRange

0.75

angan

0.75

hesda

0.74

anova

0.74

thodox

0.73

onite

0.73

Activations Density 0.000%

No Known Activations

This feature has no known activations.