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

 prepar

-0.65

OHN

-0.63

 mathemat

-0.63

pires

-0.61

 scraping

-0.60

theless

-0.60

 Macron

-0.59

boy

-0.59

azing

-0.59

 scrape

-0.59

POSITIVE LOGITS

0.83

 guiActiveUnfocused

0.83

bryce

0.81

edIn

0.75

natureconservancy

0.70

 guiName

0.69

DAQ

0.67

abus

0.65

eering

0.64

CVE

0.61

Activations Density 0.000%

No Known Activations

This feature has no known activations.