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

Ern

-0.86

Nun

-0.66

 Included

-0.64

 Falls

-0.64

 terminated

-0.63

 administr

-0.61

 indefinitely

-0.61

 briefed

-0.61

 represented

-0.60

 Uran

-0.60

POSITIVE LOGITS

IVES

0.88

ADE

0.86

efe

0.85

WAYS

0.82

vati

0.76

IFE

0.74

akings

0.74

iffin

0.73

ONES

0.72

ackle

0.72

Activations Density 0.000%

No Known Activations

This feature has no known activations.