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

BuyableInstoreAndOnline

-1.04

MQ

-0.75

xxx

-0.72

ĪĴ

-0.72

Writer

-0.71

edin

-0.69

breaks

-0.69

ãĤ´ãĥ³

-0.66

onday

-0.66

quit

-0.66

POSITIVE LOGITS

ortion

0.72

orting

0.66

emon

0.62

omen

0.62

ious

0.61

astical

0.59

ifice

0.59

igious

0.58

elligent

0.58

nder

0.58

Activations Density 0.000%

No Known Activations

This feature has no known activations.