INDEX

Explanations

No Explanations Found

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

WER

-0.77

wait

-0.71

lus

-0.68

DN

-0.66

 SERV

-0.66

 TEAM

-0.65

lda

-0.65

url

-0.65

emi

-0.65

orns

-0.65

POSITIVE LOGITS

ominium

0.74

ricting

0.66

lements

0.65

 Construction

0.65

 Brunswick

0.64

annis

0.64

 infused

0.63

 Nadu

0.63

ugal

0.62

 Azure

0.62

Activations Density 0.000%

No Known Activations

This feature has no known activations.