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

-0.88

 Austral

-0.74

ãĥĨãĤ£

-0.70

Irish

-0.67

 Shape

-0.67

 Donation

-0.67

workshop

-0.65

>[

-0.64

 Fellowship

-0.64

youtu

-0.64

POSITIVE LOGITS

 diplom

0.73

 omin

0.69

 importantly

0.69

vac

0.66

 flourish

0.64

 blur

0.62

 carp

0.62

 impress

0.62

Mom

0.62

 curric

0.61

Activations Density 0.000%

No Known Activations

This feature has no known activations.