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

nih

-0.71

gage

-0.70

athed

-0.70

bones

-0.67

estinal

-0.65

cig

-0.65

brates

-0.64

irez

-0.63

crew

-0.63

OSS

-0.63

POSITIVE LOGITS

olitan

0.70

 Discord

0.68

 Conclusion

0.66

 Extrem

0.66

 Prosper

0.62

 Petr

0.61

 Noir

0.60

 Kore

0.60

 Magnus

0.59

 instability

0.58

Activations Density 0.000%

No Known Activations

This feature has no known activations.