INDEX

Explanations

No Explanations Found

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

vu

-0.67

 durable

-0.66

 visc

-0.65

terday

-0.61

 choking

-0.61

 calming

-0.59

uchs

-0.58

 easing

-0.58

 graduating

-0.57

 decriminal

-0.56

POSITIVE LOGITS

mented

0.72

ventus

0.70

 Canary

0.69

irlf

0.68

lyak

0.68

 Wonderland

0.65

dun

0.65

FML

0.65

ittens

0.65

bench

0.64

Activations Density 0.000%

No Known Activations

This feature has no known activations.