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

 billed

-0.67

bender

-0.63

 lawn

-0.62

 welcomed

-0.62

each

-0.61

Bou

-0.61

Nug

-0.60

ean

-0.59

 Ribbon

-0.59

 billing

-0.59

POSITIVE LOGITS

acca

0.88

ventus

0.84

iets

0.80

apo

0.80

acas

0.80

ixel

0.79

zek

0.78

ivia

0.77

isoft

0.76

abwe

0.75

Activations Density 0.000%

No Known Activations

This feature has no known activations.