INDEX

Explanations

attends to various tokens marked by unique formats or characters from subsequent tokens

oai_attention-head · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GEMMA-2-2B @ 10-gemmascope-att-16k

Configuration

google/gemma-scope-2b-pt-att/layer_10/width_16k/average_l0_70

Prompts (Dashboard)

36,864 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

16,384

Data Type

float32

Hook Name

blocks.10.attn.hook_z

Hook Layer

Architecture

jumprelu

Context Size

1,024

Dataset

monology/pile-uncopyrighted

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Head Attr Weights

0:0.06

1:0.15

2:0.10

3:0.08

4:0.27

5:0.09

6:0.14

7:0.08

Negative Logits

帖最后由

-0.30

 оригіналу

-0.25

ený

-0.23

IVEREF

-0.23

További

-0.21

咤

-0.21

務省

-0.21

彼

-0.20

mif

-0.20

ycznego

-0.20

POSITIVE LOGITS

 Audiodateien

0.34

0.32

?>">

0.31

 surla

0.30

calaure

0.30

sizeCache

0.29

ToScroll

0.29

➌

0.29

 sẻ

0.28

 NDEBUG

0.28

Activations Density 0.031%

attends to various tokens marked by unique formats or characters from subsequent tokens

No Comments

No Known Activations