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

erva

-0.80

erv

-0.79

=~

-0.79

opian

-0.70

ï¸

-0.68

 Â¥

-0.66

eway

-0.64

rust

-0.64

ulet

-0.64

Ctrl

-0.63

POSITIVE LOGITS

 Cosponsors

0.85

 Stories

0.71

aughs

0.70

 churches

0.70

 Principles

0.69

 Actions

0.68

 Reports

0.68

Dy

0.68

 Types

0.67

Ips

0.66

Activations Density 0.000%

No Known Activations

This feature has no known activations.