INDEX

Explanations

No Explanations Found

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

efficients

-0.77

fman

-0.73

elman

-0.71

egu

-0.70

 Quincy

-0.69

oki

-0.69

ante

-0.68

iren

-0.67

ierrez

-0.67

reau

-0.66

POSITIVE LOGITS

regon

0.69

CN

0.65

 padd

0.63

 strip

0.63

 merge

0.62

 Soccer

0.59

 sport

0.59

 lett

0.59

 fencing

0.58

 nort

0.58

Activations Density 0.000%

No Known Activations

This feature has no known activations.