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

 Igor

-0.63

OUGH

-0.63

 Kuro

-0.62

éĹĺ

-0.61

tha

-0.60

Ku

-0.58

ItemTracker

-0.58

Â´

-0.58

stuff

-0.58

 bailed

-0.58

POSITIVE LOGITS

tone

0.81

 sexes

0.78

ults

0.73

 hasht

0.71

 equally

0.68

nets

0.68

 Virtue

0.66

college

0.65

nard

0.63

ilde

0.63

Activations Density 0.000%

No Known Activations

This feature has no known activations.