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

 Garg

-0.70

tam

-0.68

Ley

-0.64

-0.63

mel

-0.62

 fumes

-0.61

Gau

-0.60

gul

-0.60

Gaza

-0.59

ufact

-0.59

POSITIVE LOGITS

ï¸ı

0.81

Ã±o

0.77

gregation

0.75

paces

0.73

cius

0.72

ym

0.71

bledon

0.69

Fine

0.68

weekly

0.68

roth

0.67

Activations Density 0.000%

No Known Activations

This feature has no known activations.