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

personal

-0.79

ynthesis

-0.74

bug

-0.73

İĭ

-0.73

gian

-0.73

tone

-0.72

angel

-0.72

cedented

-0.71

drawn

-0.70

handled

-0.69

POSITIVE LOGITS

 Moran

0.74

 Ichigo

0.73

Zur

0.72

 Norris

0.68

 Mald

0.68

 plaster

0.66

 Archdemon

0.66

 Corpus

0.65

Sob

0.64

 fort

0.64

Activations Density 0.000%

No Known Activations

This feature has no known activations.