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

accompan

-0.89

xit

-0.74

 contrace

-0.70

morrow

-0.70

merce

-0.70

occup

-0.70

 toget

-0.69

ouk

-0.67

termination

-0.65

 Charter

-0.65

POSITIVE LOGITS

NP

0.70

rote

0.68

urus

0.68

asm

0.66

URE

0.66

ure

0.65

ONSORED

0.64

 Blend

0.61

LP

0.60

vine

0.60

Activations Density 0.000%

No Known Activations

This feature has no known activations.