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

Legendary

-0.75

 religiously

-0.71

humane

-0.63

skill

-0.61

 Heroic

-0.60

 haste

-0.60

Demon

-0.59

>]

-0.58

Magikarp

-0.58

Champ

-0.58

POSITIVE LOGITS

earances

0.82

uctions

0.77

uction

0.73

ickers

0.70

 nodd

0.69

uity

0.67

odynamics

0.67

 rebate

0.67

iddler

0.66

rollers

0.66

Activations Density 0.000%

No Known Activations

This feature has no known activations.