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

oglu

-0.73

cher

-0.67

 rust

-0.66

omach

-0.63

MpServer

-0.62

raught

-0.62

elia

-0.61

chers

-0.61

 seam

-0.61

ayers

-0.61

POSITIVE LOGITS

 Neon

0.78

atown

0.75

 Nope

0.69

schild

0.68

 Pengu

0.67

Yen

0.67

peria

0.66

 Vaughn

0.66

 Trent

0.65

 Marijuana

0.65

Activations Density 0.000%

No Known Activations

This feature has no known activations.