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

lehem

-0.81

Chicken

-0.70

Crime

-0.68

CLOSE

-0.68

__

-0.68

Sand

-0.67

NK

-0.67

Quote

-0.67

Vill

-0.66

Thumbnail

-0.66

POSITIVE LOGITS

 sweats

0.74

 slee

0.68

enser

0.67

ocrine

0.67

 booklet

0.65

 exempted

0.65

 endowed

0.64

 acquaint

0.63

 eBook

0.61

 bureaucr

0.60

Activations Density 0.000%

No Known Activations

This feature has no known activations.