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

thur

-0.72

 mediation

-0.67

 independence

-0.66

 renovation

-0.66

beit

-0.64

 renovations

-0.64

gur

-0.62

 impunity

-0.61

mun

-0.61

æĪ¦

-0.60

POSITIVE LOGITS

::::::::

0.71

urtles

0.70

ocity

0.68

Pg

0.68

iker

0.68

aughs

0.68

agascar

0.66

ÅĤ

0.66

osi

0.66

 Defin

0.66

Activations Density 0.000%

No Known Activations

This feature has no known activations.