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

 theoret

-0.72

©¶æ

-0.63

PIN

-0.60

onal

-0.60

 opposes

-0.60

SPONSORED

-0.59

can

-0.59

 abide

-0.59

ivist

-0.59

 borne

-0.57

POSITIVE LOGITS

egu

0.73

 Carbuncle

0.71

ptions

0.71

Pax

0.70

ogg

0.69

addon

0.67

ellig

0.67

 Machines

0.67

 Rept

0.67

Nid

0.66

Activations Density 0.000%

No Known Activations

This feature has no known activations.