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

 extrad

-0.81

 Democr

-0.75

 FANTASY

-0.72

 Corvette

-0.70

 acknow

-0.66

 criminally

-0.66

 intervening

-0.65

 diplom

-0.65

 Interstellar

-0.64

 Draco

-0.61

POSITIVE LOGITS

sticks

0.64

iasis

0.63

Fail

0.62

=]

0.62

 Eating

0.61

 burn

0.61

ĺħ

0.60

 Needs

0.59

Soc

0.59

sg

0.58

Activations Density 0.000%

No Known Activations

This feature has no known activations.