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

acebook

-0.81

SpaceEngineers

-0.69

foundland

-0.66

mares

-0.66

é¾

-0.65

ikuman

-0.65

 Retrieved

-0.64

ETHOD

-0.63

 intentions

-0.62

 Nanto

-0.61

POSITIVE LOGITS

thy

0.72

leans

0.68

 Laksh

0.64

NetMessage

0.64

GR

0.64

ivable

0.63

lean

0.62

 Reps

0.61

ios

0.60

 strain

0.60

Activations Density 0.000%

No Known Activations

This feature has no known activations.