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

kefeller

-0.74

ific

-0.73

lav

-0.71

ipel

-0.66

 Rolls

-0.63

 largeDownload

-0.62

pend

-0.62

 Nanto

-0.60

swe

-0.59

ificate

-0.59

POSITIVE LOGITS

 while

0.83

regate

0.70

$$$$

0.68

 Akron

0.61

 Horizon

0.61

 Cros

0.59

 Bear

0.59

 Grizz

0.59

wine

0.59

ctor

0.58

Activations Density 0.000%

No Known Activations

This feature has no known activations.