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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 COURT

-0.84

ourse

-0.68

uckland

-0.68

oner

-0.67

 awoken

-0.65

essage

-0.65

pex

-0.63

iple

-0.63

 overcrowd

-0.62

 traged

-0.61

POSITIVE LOGITS

 compared

0.95

parts

0.87

ãģķ

0.68

lehem

0.63

cond

0.63

Morning

0.63

Money

0.63

ivari

0.63

rote

0.62

Activations Density 0.000%

No Known Activations

This feature has no known activations.