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

veyard

-0.70

Lago

-0.70

ndra

-0.67

 crocod

-0.66

maxwell

-0.66

EStream

-0.64

irs

-0.64

redo

-0.63

Cu

-0.63

erences

-0.62

POSITIVE LOGITS

iam

0.78

izo

0.64

imus

0.61

lan

0.61

zyk

0.61

iversal

0.60

 Virtual

0.60

hend

0.60

ize

0.59

iz

0.58

Activations Density 0.000%

No Known Activations

This feature has no known activations.