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

votes

-0.86

aunder

-0.72

cffffcc

-0.69

ãĥ¤

-0.69

letters

-0.66

quart

-0.65

 certs

-0.65

BALL

-0.64

AUT

-0.62

OGR

-0.61

POSITIVE LOGITS

©¶æ

0.82

 Called

0.72

iday

0.69

ighed

0.62

rist

0.61

atered

0.61

ħĭ

0.59

 unchecked

0.59

feat

0.58

 abiding

0.58

Activations Density 0.000%

No Known Activations

This feature has no known activations.