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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 ingred

-0.83

 encount

-0.72

 advoc

-0.67

 Afric

-0.67

 Thrones

-0.65

 Lomb

-0.64

clud

-0.63

ore

-0.62

 Icelandic

-0.62

 therap

-0.61

POSITIVE LOGITS

holes

0.77

0.75

leeve

0.72

å·

0.72

ãĥĺãĥ©

0.70

amsung

0.69

perature

0.68

atches

0.68

atis

0.67

eton

0.67

Activations Density 0.000%

No Known Activations

This feature has no known activations.