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

pps

-0.69

DN

-0.64

seless

-0.64

ypes

-0.62

liner

-0.61

Pak

-0.60

lining

-0.59

leen

-0.59

 Upgrade

-0.58

orough

-0.57

POSITIVE LOGITS

usterity

0.75

ersive

0.74

ipal

0.74

OTOS

0.73

[|

0.66

duty

0.65

dylib

0.65

 borne

0.64

 Mandela

0.64

+++

0.62

Activations Density 0.000%

No Known Activations

This feature has no known activations.