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

oute

-0.81

rift

-0.77

 millenn

-0.74

 suspic

-0.74

si

-0.73

Rai

-0.73

edition

-0.72

ija

-0.71

senal

-0.69

 arrang

-0.69

POSITIVE LOGITS

estead

0.67

dden

0.66

TBD

0.65

========

0.65

Bed

0.64

 Vegas

0.61

bernatorial

0.61

 Unle

0.61

 guiName

0.59

ftime

0.59

Activations Density 0.000%

No Known Activations

This feature has no known activations.