INDEX

Explanations

No Explanations Found

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

esson

-0.82

hement

-0.77

usha

-0.74

ridor

-0.70

dinand

-0.70

oru

-0.70

ebted

-0.69

reddits

-0.68

 droid

-0.68

crew

-0.67

POSITIVE LOGITS

idia

0.82

ieth

0.70

ODE

0.66

Plug

0.64

 authenticated

0.63

IOR

0.62

é¾į

0.62

vable

0.61

lit

0.60

 Came

0.60

Activations Density 0.000%

No Known Activations

This feature has no known activations.