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

Paw

-0.78

ilitary

-0.67

awar

-0.67

uala

-0.67

 Pipeline

-0.65

qv

-0.65

itsch

-0.64

 gobl

-0.64

Bow

-0.62

merga

-0.62

POSITIVE LOGITS

spin

0.70

sheets

0.70

sect

0.70

tails

0.69

leep

0.68

oi

0.67

cember

0.66

eering

0.65

rator

0.64

 Bleach

0.63

Activations Density 0.000%

No Known Activations

This feature has no known activations.