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

grain

-0.75

soType

-0.72

ishable

-0.69

 hast

-0.69

agy

-0.69

raft

-0.69

 corrid

-0.68

ŃĶ

-0.65

uten

-0.64

terness

-0.64

POSITIVE LOGITS

 However

0.85

 Whereas

0.83

 Typically

0.79

 Presumably

0.79

archives

0.78

 Flavoring

0.77

 Luckily

0.77

 Unfortunately

0.77

 Alternatively

0.76

 Nonetheless

0.76

Activations Density 0.000%

No Known Activations

This feature has no known activations.