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

 nomine

-0.80

umbnails

-0.79

eva

-0.77

egu

-0.73

ogo

-0.69

oho

-0.69

odka

-0.69

 contagious

-0.67

utterstock

-0.66

orously

-0.65

POSITIVE LOGITS

heses

0.85

 Compass

0.81

CAST

0.80

casts

0.67

catentry

0.66

HUD

0.66

[&

0.65

 Riders

0.64

 Voltage

0.64

 GEAR

0.62

Activations Density 0.000%

No Known Activations

This feature has no known activations.