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

qus

-0.73

esm

-0.71

eworld

-0.71

fml

-0.70

cyclopedia

-0.69

luaj

-0.67

eria

-0.67

ESCO

-0.65

ricia

-0.65

kefeller

-0.64

POSITIVE LOGITS

 Wrap

0.68

 Postal

0.68

 absentee

0.66

 clipboard

0.66

votes

0.65

 postal

0.64

hots

0.64

 executions

0.63

shots

0.63

election

0.62

Activations Density 0.000%

No Known Activations

This feature has no known activations.