INDEX

Explanations

No Explanations Found

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

 climb

-0.65

erva

-0.64

doors

-0.63

nee

-0.62

nel

-0.62

ARP

-0.62

 platform

-0.62

Pe

-0.62

ariat

-0.61

 Rapp

-0.60

POSITIVE LOGITS

theless

0.86

sembly

0.74

 misunder

0.74

ecause

0.71

 Saints

0.68

oun

0.68

inging

0.67

²¾

0.67

unit

0.65

bye

0.64

Activations Density 0.000%

No Known Activations

This feature has no known activations.