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

hra

-0.80

thy

-0.77

veyard

-0.73

reys

-0.73

 tongues

-0.70

 praises

-0.69

 ignition

-0.69

eth

-0.69

 Egyptians

-0.68

 Hers

-0.67

POSITIVE LOGITS

è¡

0.78

é¾įåĸļå£«

0.73

è£ħ

0.71

Optional

0.69

visory

0.69

VERTISEMENT

0.69

visors

0.69

intendent

0.68

MpServer

0.67

Variable

0.67

Activations Density 0.000%

No Known Activations

This feature has no known activations.