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

 Revelations

-0.73

VAT

-0.72

Ink

-0.69

 Zeit

-0.68

 Ascension

-0.66

 Horizons

-0.66

ancial

-0.65

 Communications

-0.65

 EntityItem

-0.65

 Covenant

-0.64

POSITIVE LOGITS

acher

0.85

kee

0.85

hur

0.84

uncle

0.79

pees

0.77

mol

0.75

pee

0.73

upe

0.71

old

0.71

ãĥĦ

0.70

Activations Density 0.000%

No Known Activations

This feature has no known activations.