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

 Dominion

-0.77

 Robertson

-0.68

 Horizon

-0.67

 Samar

-0.66

ories

-0.64

 Canaver

-0.60

Kot

-0.58

 Kirk

-0.58

KNOWN

-0.58

 Larson

-0.57

POSITIVE LOGITS

ele

0.83

fee

0.80

STATE

0.79

laws

0.74

law

0.74

 weeds

0.72

Bloom

0.69

ois

0.69

law

0.68

 state

0.67

Activations Density 0.000%

No Known Activations

This feature has no known activations.