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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

mers

-0.67

untled

-0.67

 reconstructed

-0.66

 impress

-0.65

 motor

-0.63

kered

-0.63

 behav

-0.63

 cannon

-0.60

 resil

-0.60

 todd

-0.60

POSITIVE LOGITS

iPhone

0.80

SourceFile

0.74

onite

0.71

Issue

0.68

Choice

0.68

Earth

0.68

uggets

0.68

zn

0.68

lishes

0.67

foundland

0.66

Activations Density 0.000%

No Known Activations

This feature has no known activations.