INDEX

Explanations

No Explanations Found

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

chery

-0.79

aukee

-0.71

oral

-0.69

olson

-0.66

)=(

-0.66

ogene

-0.64

uni

-0.63

edIn

-0.63

kef

-0.62

 Yemeni

-0.62

POSITIVE LOGITS

ince

0.65

Sco

0.61

matic

0.60

 plaster

0.58

isse

0.57

shoot

0.56

 entertain

0.56

 Pocket

0.55

uctor

0.55

 dashboard

0.55

Activations Density 0.000%

No Known Activations

This feature has no known activations.