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

 Whilst

-0.75

 withd

-0.71

usercontent

-0.70

 âī¡

-0.67

 âĩ

-0.64

 hither

-0.64

(<

-0.64

 Casting

-0.63

 Featuring

-0.62

 âĸº

-0.61

POSITIVE LOGITS

 those

1.28

 that

1.06

the

0.91

 THAT

0.79

its

0.77

those

0.76

that

0.70

 their

0.70

0.69

 anyone

0.68

Activations Density 0.000%

No Known Activations

This feature has no known activations.