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

Ĥ

-0.72

Capture

-0.72

FS

-0.70

 Cunningham

-0.70

 Channel

-0.68

han

-0.66

é¾įåĸļå£«

-0.66

NC

-0.64

Channel

-0.64

·

-0.63

POSITIVE LOGITS

 streng

0.78

rul

0.77

 acknow

0.76

 conduc

0.73

 tomorrow

0.73

 hashing

0.72

 constitu

0.71

olkien

0.71

 ingred

0.69

 Ukrain

0.69

Activations Density 0.000%

No Known Activations

This feature has no known activations.