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

finding

-0.82

bowl

-0.81

atoes

-0.78

assium

-0.76

food

-0.74

bread

-0.71

umbers

-0.70

boys

-0.69

iencies

-0.68

 Sodium

-0.68

POSITIVE LOGITS

)."

0.75

)"

0.72

ACP

0.72

.).

0.68

 acknow

0.67

UTC

0.67

Syn

0.66

 Mast

0.65

!)

0.65

îĢ

0.62

Activations Density 0.000%

No Known Activations

This feature has no known activations.