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

mma

-0.75

ombat

-0.70

tackle

-0.69

ifix

-0.68

capital

-0.67

phal

-0.66

xe

-0.66

xes

-0.65

onal

-0.65

 Boko

-0.63

POSITIVE LOGITS

âĸĪâĸĪâĸĪâĸĪ

0.68

 hats

0.63

 FINAL

0.62

ARC

0.61

rawdownloadcloneembedreportprint

0.61

iets

0.61

entin

0.59

 antioxid

0.59

æĪ¦

0.58

 Seal

0.58

Activations Density 0.000%

No Known Activations

This feature has no known activations.