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

interstitial

-0.88

Ħ¢

-0.69

resist

-0.65

index

-0.65

arta

-0.65

¶ħ

-0.64

nings

-0.64

kus

-0.64

ãĥ¼ãĤ¯

-0.64

steel

-0.63

POSITIVE LOGITS

 lone

0.99

 Galile

0.72

 Fern

0.66

 afar

0.62

 delegated

0.61

 Nest

0.61

elia

0.60

Constructed

0.60

 Leap

0.59

oit

0.58

Activations Density 0.000%

No Known Activations

This feature has no known activations.