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

 Twist

-0.70

 lows

-0.65

 Harvest

-0.64

 Imper

-0.63

 Palest

-0.60

 Tail

-0.60

Sup

-0.59

itia

-0.59

 CLSID

-0.59

 Tale

-0.58

POSITIVE LOGITS

advertising

2.04

 alive

0.92

Dex

0.67

Benz

0.66

CLOSE

0.64

vous

0.64

merce

0.63

sic

0.62

moving

0.60

chrome

0.59

Activations Density 0.000%

No Known Activations

This feature has no known activations.