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

ulla

-0.72

hots

-0.71

erey

-0.71

resy

-0.69

Preview

-0.67

haw

-0.66

Oro

-0.65

Common

-0.64

BIL

-0.64

 Booker

-0.63

POSITIVE LOGITS

employment

0.67

 empires

0.63

shit

0.62

 punch

0.61

hook

0.61

 empowerment

0.61

EN

0.60

ryce

0.60

luster

0.59

uminati

0.58

Activations Density 0.000%

No Known Activations

This feature has no known activations.