INDEX

Explanations

No Explanations Found

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

 occurrences

-0.68

oms

-0.65

 Eleven

-0.65

 Mahar

-0.61

 strains

-0.59

 Lines

-0.59

ograms

-0.58

 Doodle

-0.57

 contenders

-0.56

 Mavericks

-0.56

POSITIVE LOGITS

pite

0.78

bane

0.76

oa

0.68

DCS

0.68

uracy

0.64

ortium

0.63

RG

0.63

hip

0.63

byter

0.62

RPG

0.61

Activations Density 0.000%

No Known Activations

This feature has no known activations.