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

gas

-0.73

 converter

-0.69

vernment

-0.68

team

-0.67

anged

-0.63

panel

-0.63

billion

-0.62

 totaled

-0.61

packages

-0.61

clusively

-0.61

POSITIVE LOGITS

 Flavoring

0.74

Ire

0.71

LED

0.70

Ô

0.69

INA

0.69

ikk

0.67

éĸ

0.65

layer

0.65

edIn

0.65

uania

0.64

Activations Density 0.000%

No Known Activations

This feature has no known activations.