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

hower

-0.93

issance

-0.76

itta

-0.74

ICLE

-0.71

 fielded

-0.68

agher

-0.65

ski

-0.64

instein

-0.64

naissance

-0.64

imeter

-0.62

POSITIVE LOGITS

otiation

0.73

oids

0.69

Vil

0.67

 wards

0.66

alyses

0.66

Dru

0.65

oid

0.63

 Stones

0.62

 chromosomes

0.62

ocard

0.61

Activations Density 0.000%

No Known Activations

This feature has no known activations.