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

ulum

-0.73

 ancest

-0.68

 technician

-0.66

 immobil

-0.64

 Instruct

-0.61

orned

-0.61

IDENT

-0.61

 Technician

-0.60

obil

-0.58

ription

-0.58

POSITIVE LOGITS

tracks

0.73

games

0.73

fun

0.72

faces

0.71

pps

0.70

follow

0.70

few

0.68

coins

0.67

cles

0.67

++

0.66

Activations Density 0.000%

No Known Activations

This feature has no known activations.