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

Rai

-0.78

lehem

-0.70

igun

-0.70

 corrid

-0.69

 enthusi

-0.69

ppo

-0.68

zbollah

-0.68

zona

-0.67

ombo

-0.67

bom

-0.65

POSITIVE LOGITS

owed

0.83

hold

0.68

backs

0.66

igned

0.62

 Furn

0.61

 Dull

0.61

iors

0.60

aken

0.60

0.59

atical

0.58

Activations Density 0.000%

No Known Activations

This feature has no known activations.