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

wich

-0.69

ved

-0.68

icial

-0.66

 swall

-0.63

 gulf

-0.63

 guiActiveUn

-0.62

otom

-0.60

 CLIENT

-0.60

icter

-0.59

cf

-0.59

POSITIVE LOGITS

MpServer

0.72

aterasu

0.68

iggs

0.66

alted

0.66

IUM

0.65

icum

0.62

emetery

0.62

SPEC

0.61

Bio

0.61

uria

0.58

Activations Density 0.000%

No Known Activations

This feature has no known activations.