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

chuk

-0.71

å£«

-0.70

 Nich

-0.66

enko

-0.66

 Howell

-0.66

ãĥ¯ãĥ³

-0.65

cloth

-0.65

 Tata

-0.65

verty

-0.64

cu

-0.64

POSITIVE LOGITS

ropolitan

0.73

 filing

0.72

osponsors

0.71

urst

0.70

isconsin

0.70

urate

0.70

hibit

0.69

uras

0.68

reon

0.65

MpServer

0.65

Activations Density 0.000%

No Known Activations

This feature has no known activations.