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

 rabbits

-0.68

uca

-0.66

 Zombies

-0.66

terday

-0.65

Vi

-0.65

âĢ

-0.65

soType

-0.64

sew

-0.62

Wiz

-0.59

 downt

-0.59

POSITIVE LOGITS

angular

0.92

pine

0.90

flex

0.85

amic

0.79

accompan

0.79

ixt

0.77

lished

0.77

andon

0.77

restricted

0.76

reve

0.76

Activations Density 0.000%

No Known Activations

This feature has no known activations.