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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 prevailed

-0.74

 theorem

-0.66

 hump

-0.62

MSG

-0.62

 Polo

-0.62

odan

-0.61

mma

-0.60

 ignor

-0.60

@@@@@@@@

-0.60

 cous

-0.59

POSITIVE LOGITS

leaf

0.89

mobi

0.85

testing

0.75

rough

0.73

hers

0.72

places

0.65

ranged

0.65

serving

0.65

ãĤ£

0.64

ãĤ©

0.63

Activations Density 0.000%

No Known Activations

This feature has no known activations.