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

gap

-0.81

iven

-0.80

agra

-0.74

inguishable

-0.72

anchester

-0.71

psons

-0.69

encer

-0.69

vation

-0.68

âĸĪâĸĪâĸĪâĸĪâĸĪâĸĪâĸĪâĸĪ

-0.68

icter

-0.67

POSITIVE LOGITS

 email

0.84

Tex

0.69

mix

0.66

Editor

0.62

 edit

0.62

 bump

0.62

UL

0.61

dip

0.58

 Ulster

0.58

 Forth

0.58

Activations Density 0.000%

No Known Activations

This feature has no known activations.