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

ruciating

-0.67

UGE

-0.66

âĻ¥

-0.64

Som

-0.63

 colon

-0.63

 peac

-0.62

eous

-0.62

 phosphate

-0.62

wre

-0.60

rite

-0.59

POSITIVE LOGITS

ãĥ¼ãĥĨãĤ£

0.85

 Gate

0.73

DragonMagazine

0.69

}}}

0.67

Aad

0.67

 Machines

0.66

 Wire

0.65

0.64

 Graphics

0.63

sbm

0.62

Activations Density 0.000%

No Known Activations

This feature has no known activations.