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

Applications

-0.77

 denomin

-0.69

 deviations

-0.67

Action

-0.65

 conversions

-0.64

iott

-0.64

Administ

-0.63

 Physicians

-0.62

 litres

-0.62

letters

-0.62

POSITIVE LOGITS

cano

0.80

warm

0.80

dden

0.74

afort

0.73

avin

0.71

 Lennon

0.70

ĸļ

0.67

osaurus

0.66

inson

0.66

chin

0.65

Activations Density 0.000%

No Known Activations

This feature has no known activations.