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

sing

-0.69

Bei

-0.69

æĪ¦

-0.68

blance

-0.67

bler

-0.67

achy

-0.66

ART

-0.65

ogly

-0.65

ŃĶ

-0.65

bling

-0.64

POSITIVE LOGITS

aminer

0.92

actionDate

0.89

 happ

0.74

aughs

0.64

 disqualified

0.62

utterstock

0.62

othal

0.61

 Gazette

0.60

 Client

0.59

 appre

0.58

Activations Density 0.000%

No Known Activations

This feature has no known activations.