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

ãĥĬ

-0.69

lip

-0.66

dyn

-0.64

 Instruction

-0.63

è¦ļéĨĴ

-0.63

nat

-0.63

english

-0.61

fy

-0.59

lit

-0.59

gin

-0.58

POSITIVE LOGITS

urry

0.68

 Watson

0.65

ieri

0.64

emer

0.64

arij

0.61

assadors

0.61

resistant

0.60

 Whis

0.60

 Emer

0.60

 frenzy

0.59

Activations Density 0.000%

No Known Activations

This feature has no known activations.