INDEX

Explanations

No Explanations Found

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

utan

-0.75

kov

-0.75

rage

-0.74

kan

-0.74

vell

-0.73

escal

-0.73

ighth

-0.73

Subject

-0.72

abuse

-0.71

}:

-0.70

POSITIVE LOGITS

 Kobe

0.82

 pastry

0.76

 DRAG

0.75

 Takeru

0.75

 metro

0.74

SAP

0.72

 WARN

0.72

CLR

0.70

 Hawai

0.70

 Tsukuyomi

0.69

Activations Density 0.000%

No Known Activations

This feature has no known activations.