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

auga

-0.93

iosity

-0.81

iasm

-0.80

gow

-0.76

;;;;;;;;;;;;

-0.75

igmatic

-0.75

tions

-0.73

yssey

-0.72

DragonMagazine

-0.72

aurus

-0.71

POSITIVE LOGITS

 board

0.79

 Topic

0.73

 fiat

0.71

 mistake

0.71

 avoid

0.70

 boost

0.68

 cram

0.67

 virtue

0.67

 Presidents

0.67

 whoever

0.66

Activations Density 0.000%

No Known Activations

This feature has no known activations.