INDEX

Explanations

No Explanations Found

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

SHIP

-0.86

FB

-0.74

ilk

-0.74

UU

-0.73

retty

-0.72

ADRA

-0.71

Demand

-0.71

ittal

-0.70

aspberry

-0.69

enza

-0.69

POSITIVE LOGITS

 Pandora

0.76

 veins

0.69

raph

0.68

Yad

0.65

 Tags

0.64

 transl

0.63

 paced

0.61

 imaginable

0.59

 appraisal

0.58

 Eden

0.58

Activations Density 0.000%

No Known Activations

This feature has no known activations.