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

ensical

-0.72

erella

-0.71

redo

-0.67

oodle

-0.66

isec

-0.66

oglobin

-0.65

dule

-0.64

 cannabin

-0.63

ricular

-0.63

odox

-0.61

POSITIVE LOGITS

loving

0.68

éģ

0.64

Shipping

0.64

Pand

0.64

Eva

0.64

Tax

0.60

estern

0.60

Mart

0.60

tackle

0.59

 Lilith

0.58

Activations Density 0.000%

No Known Activations

This feature has no known activations.