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

 explo

-0.73

rawdownloadcloneembedreportprint

-0.73

nat

-0.71

isco

-0.70

brow

-0.66

 unse

-0.65

 nont

-0.65

hack

-0.63

 isEnabled

-0.62

 nonex

-0.61

POSITIVE LOGITS

æī

0.81

ãĤ·ãĥ£

0.81

ãĥ¼ãĥĨ

0.80

amins

0.80

ãĥŁ

0.77

 Brav

0.77

lihood

0.75

peria

0.75

Redditor

0.74

ãĤ©

0.72

Activations Density 0.000%

No Known Activations

This feature has no known activations.