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

izoph

-0.75

flies

-0.73

 pads

-0.67

 livest

-0.67

 colourful

-0.65

ophe

-0.65

 butterflies

-0.63

 artwork

-0.62

 laure

-0.62

ellery

-0.62

POSITIVE LOGITS

yet

0.91

yet

0.84

Hack

0.76

Yet

0.75

ãĤ°

0.74

Critical

0.73

Xi

0.72

bound

0.71

MENTS

0.70

Shift

0.68

Activations Density 0.000%

No Known Activations

This feature has no known activations.