INDEX

Explanations

No Explanations Found

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

 Anniversary

-0.76

achev

-0.74

 Massacre

-0.70

ISO

-0.62

 hello

-0.62

 Pastebin

-0.61

 closure

-0.60

 surprise

-0.60

 Miliband

-0.60

 Rend

-0.59

POSITIVE LOGITS

rust

0.85

rising

0.71

rights

0.71

è¦ļéĨĴ

0.70

 borne

0.69

hes

0.68

localhost

0.67

 propelled

0.67

gat

0.67

RT

0.67

Activations Density 0.000%

No Known Activations

This feature has no known activations.