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

igr

-0.71

antic

-0.67

aline

-0.67

 plunge

-0.66

ascal

-0.66

etheless

-0.65

atten

-0.64

roxy

-0.64

ĸļ

-0.64

ordial

-0.63

POSITIVE LOGITS

,,,,

0.76

NEC

0.72

tion

0.69

 Zamb

0.68

 Palest

0.68

ONS

0.68

 Females

0.66

minist

0.65

OAD

0.62

 Elias

0.61

Activations Density 0.000%

No Known Activations

This feature has no known activations.