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

pled

-0.83

ocular

-0.81

accessible

-0.79

rss

-0.78

asper

-0.75

bats

-0.74

atories

-0.69

vv

-0.69

ebook

-0.69

plates

-0.69

POSITIVE LOGITS

 Macy

0.76

 Marqu

0.71

 Warrant

0.66

 Rosa

0.66

 Royal

0.66

 report

0.64

 Smithsonian

0.63

 Mandatory

0.63

 Prel

0.62

 Papua

0.62

Activations Density 0.000%

No Known Activations

This feature has no known activations.