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

vae

-0.67

ris

-0.62

ETA

-0.59

 dedication

-0.59

 naval

-0.59

onse

-0.58

ktop

-0.58

FORMATION

-0.57

 Latin

-0.56

riers

-0.56

POSITIVE LOGITS

ĪĴ

1.05

taboola

0.87

works

0.81

maker

0.80

 Decay

0.77

forums

0.76

cdn

0.72

odder

0.69

Ļ½

0.68

#$#$

0.68

Activations Density 0.000%

No Known Activations

This feature has no known activations.