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

 Instructor

-0.74

Berry

-0.67

 Correction

-0.65

 embargo

-0.63

 Crescent

-0.63

milo

-0.63

een

-0.59

Joined

-0.58

 Harvest

-0.58

ault

-0.57

POSITIVE LOGITS

plays

0.68

agi

0.67

0.64

acters

0.64

Vill

0.62

eers

0.62

ems

0.62

onto

0.62

å§«

0.61

çīĪ

0.61

Activations Density 0.000%

No Known Activations

This feature has no known activations.