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

ilipp

-0.78

 intern

-0.70

 Hond

-0.65

otaur

-0.65

 capt

-0.63

 Everett

-0.63

HAL

-0.63

Jol

-0.63

Dul

-0.61

 Kund

-0.60

POSITIVE LOGITS

rower

0.80

river

0.76

apple

0.74

DOM

0.73

Newsletter

0.72

Market

0.70

lance

0.69

SHIP

0.68

Beer

0.67

 fortun

0.66

Activations Density 0.000%

No Known Activations

This feature has no known activations.