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

going

-0.74

polit

-0.74

PRESS

-0.72

abolic

-0.69

mos

-0.67

LET

-0.67

Rog

-0.67

Osc

-0.66

 ammonia

-0.66

cephal

-0.65

POSITIVE LOGITS

 Citizenship

0.74

 Travels

0.73

 Haas

0.69

 Dimensions

0.68

 Removal

0.67

izon

0.67

 Tags

0.67

 Storage

0.66

 Finger

0.66

 Hamilton

0.65

Activations Density 0.000%

No Known Activations

This feature has no known activations.