INDEX

Explanations

No Explanations Found

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 2-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.2.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.2.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Negative Logits

ometimes

-0.75

theless

-0.73

 impede

-0.73

rely

-0.72

 externally

-0.70

 Leilan

-0.69

paces

-0.67

ancial

-0.67

orius

-0.67

endif

-0.66

POSITIVE LOGITS

 Flor

0.73

avi

0.69

Fab

0.68

 precincts

0.67

 Balt

0.67

ips

0.67

 Gamb

0.66

 Nost

0.64

alse

0.62

Afee

0.61

Activations Density 0.000%

No Known Activations

This feature has no known activations.