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

arat

-0.80

serv

-0.72

serving

-0.70

icides

-0.70

Sold

-0.69

esses

-0.68

\\\\\\\\

-0.65

cellent

-0.65

utical

-0.64

uted

-0.64

POSITIVE LOGITS

adder

0.73

ername

0.72

 Investigative

0.69

 Programming

0.67

 programmer

0.63

ITS

0.62

 Vulkan

0.61

hent

0.60

 Federation

0.59

 Nanto

0.59

Activations Density 0.000%

No Known Activations

This feature has no known activations.