INDEX

Explanations

No Explanations Found

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

riched

-0.74

?).

-0.67

 Unloaded

-0.67

 Though

-0.67

 Since

-0.66

vale

-0.66

 Square

-0.65

 Serving

-0.64

 AMERICA

-0.64

 Comic

-0.62

POSITIVE LOGITS

 incorpor

0.82

inki

0.78

 occupancy

0.76

igi

0.72

gencies

0.70

uilt

0.69

Bs

0.67

 occup

0.67

APD

0.67

ioch

0.67

Activations Density 0.000%

No Known Activations

This feature has no known activations.