INDEX

Explanations

conjunctions and verbs related to various topics

New Auto-Interp

Top Features by Cosine Similarity

Configuration

ckkissane/attn-saes-gpt2-small-all-layers/gpt2-small_L5_Hcat_z_lr1.20e-03_l11.00e+00_ds49152_bs4096_dc1.00e-06_rsanthropic_rie25000_nr4_v9.pt

Prompts (Dashboard)

36,864 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

49,152

Data Type

float32

Hook Name

blocks.5.attn.hook_z

Hook Layer

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Activation Function

relu

Embeds

IFrame

Link

Head Attr Weights

0:0.04

1:0.02

2:0.12

3:0.07

4:0.06

5:0.05

6:0.13

7:0.08

8:0.06

9:0.06

10:0.07

11:0.19

Negative Logits

ngth

-1.99

 tactics

-1.92

ォ

-1.91

ihilation

-1.89

keye

-1.75

 Struggle

-1.68

iken

-1.65

 cohesion

-1.64

ワン

-1.62

ggle

-1.60

POSITIVE LOGITS

taboola

2.14

 Dealer

1.98

 Loans

1.93

DragonMagazine

1.91

english

1.78

Oracle

1.77

 taxable

1.73

AUD

1.73

dow

1.72

 convertible

1.68

Activations Density 0.001%

conjunctions and verbs related to various topics

No Known Activations