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

ovember

-0.91

Hug

-0.89

 Jagu

-0.81

çĶ°

-0.76

 Sutherland

-0.75

 Gael

-0.74

 caut

-0.74

ocene

-0.74

merce

-0.74

govtrack

-0.73

POSITIVE LOGITS

ze

0.75

ious

0.66

hes

0.65

ka

0.64

one

0.64

 natural

0.64

ble

0.63

":"/

0.63

 applied

0.62

aled

0.61

Activations Density 0.000%

No Known Activations

This feature has no known activations.