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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 GoPro

-0.84

 Vinyl

-0.83

 Carnage

-0.79

 Volvo

-0.78

 Pioneer

-0.77

 Garmin

-0.75

 Staples

-0.74

reth

-0.73

Peg

-0.71

 Destination

-0.70

POSITIVE LOGITS

cess

0.82

imately

0.74

conservancy

0.73

zai

0.72

 quadru

0.72

hift

0.69

 solve

0.69

 refute

0.67

uilt

0.67

auld

0.67

Activations Density 0.000%

No Known Activations

This feature has no known activations.