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

rail

-0.72

¿½

-0.67

 reckoned

-0.66

 argue

-0.63

joy

-0.63

bane

-0.61

vu

-0.61

eye

-0.59

care

-0.58

onson

-0.58

POSITIVE LOGITS

 srfAttach

0.72

dex

0.69

rawdownloadcloneembedreportprint

0.68

 Swim

0.68

ERC

0.67

 dictators

0.66

 mamm

0.65

uters

0.65

 conventions

0.64

ç

0.64

Activations Density 0.000%

No Known Activations

This feature has no known activations.