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

 predec

-0.80

 incent

-0.80

 constitu

-0.76

 conting

-0.74

 commitments

-0.72

 ancest

-0.69

 prosec

-0.69

undai

-0.66

thinkable

-0.66

ŃĶ

-0.66

POSITIVE LOGITS

ORY

0.82

 Hale

0.73

 Tart

0.72

Kon

0.70

ually

0.70

renheit

0.69

Raz

0.69

 Griffin

0.69

 Valencia

0.68

river

0.67

Activations Density 0.000%

No Known Activations

This feature has no known activations.