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

 speaker

-0.67

 honour

-0.66

 convict

-0.66

Wr

-0.65

 Advocate

-0.64

 Lump

-0.63

Ecc

-0.62

ï¸

-0.62

 Ancient

-0.62

 Sovere

-0.61

POSITIVE LOGITS

ahime

0.92

igree

0.89

ensen

0.86

aneers

0.82

yrinth

0.81

erella

0.80

yahoo

0.79

merce

0.79

imi

0.78

anya

0.78

Activations Density 0.000%

No Known Activations

This feature has no known activations.