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

adobe

-0.80

rier

-0.73

rums

-0.70

 Grimm

-0.70

phabet

-0.70

 poorly

-0.69

ashtra

-0.66

pell

-0.66

 inadequ

-0.65

ajor

-0.65

POSITIVE LOGITS

 Ethics

0.83

isites

0.73

VIEW

0.68

Call

0.66

Books

0.65

 Becker

0.65

0.64

Pi

0.64

 Parties

0.63

GH

0.63

Activations Density 0.000%

No Known Activations

This feature has no known activations.