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

 shel

-0.64

 shelves

-0.63

sets

-0.62

ouses

-0.61

offs

-0.61

ggles

-0.61

 Helsinki

-0.60

weights

-0.59

 grad

-0.59

 regimes

-0.59

POSITIVE LOGITS

SIGN

0.87

ocument

0.79

éĹĺ

0.77

 externalToEVAOnly

0.77

Reloaded

0.77

undle

0.75

 Crusade

0.73

ĳå£«

0.72

fect

0.71

ierre

0.71

Activations Density 0.000%

No Known Activations

This feature has no known activations.