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

 stitches

-0.82

 Olsen

-0.77

mallow

-0.72

Lov

-0.71

uel

-0.70

 Walt

-0.70

 Rowe

-0.69

icz

-0.66

ãĥ¼ãĥĨ

-0.66

 Drag

-0.66

POSITIVE LOGITS

Asia

0.82

 Empires

0.80

 Nadu

0.77

 charism

0.76

_-

0.76

=]

0.73

ornings

0.70

¥µ

0.69

Gate

0.69

Pakistan

0.68

Activations Density 0.000%

No Known Activations

This feature has no known activations.