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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

ipedia

-0.75

MSG

-0.72

Nex

-0.69

 Bethlehem

-0.68

 Staples

-0.68

 abbrevi

-0.68

 Practices

-0.65

 Zucker

-0.64

 Lists

-0.64

 Topics

-0.64

POSITIVE LOGITS

inson

0.83

itton

0.82

ilipp

0.82

peror

0.82

ory

0.81

ourke

0.80

hedral

0.79

opter

0.78

orically

0.75

itals

0.74

Activations Density 0.000%

No Known Activations

This feature has no known activations.