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

paio

-0.88

-0.75

atech

-0.74

ubuntu

-0.73

interstitial

-0.70

API

-0.67

²¾

-0.67

odus

-0.66

alia

-0.66

andowski

-0.65

POSITIVE LOGITS

Nib

0.73

 nipples

0.65

ATHER

0.64

 stret

0.62

 originally

0.62

 tide

0.62

 strain

0.61

 dissatisfied

0.61

Iw

0.61

 wires

0.59

Activations Density 0.000%

No Known Activations

This feature has no known activations.