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

position

-0.74

tro

-0.68

ModLoader

-0.66

itus

-0.65

UES

-0.63

irmation

-0.62

encies

-0.62

Ott

-0.61

unt

-0.61

intention

-0.60

POSITIVE LOGITS

ä»£

0.79

finder

0.74

ĸļ

0.71

atown

0.68

 Parenthood

0.67

leys

0.64

 Reloaded

0.64

amera

0.62

esan

0.62

akeru

0.61

Activations Density 0.000%

No Known Activations

This feature has no known activations.