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

tip

-0.77

psey

-0.76

iat

-0.74

atform

-0.72

iens

-0.71

 Bust

-0.71

phalt

-0.71

Ãį

-0.70

uit

-0.66

 Drum

-0.65

POSITIVE LOGITS

 reading

1.02

 READ

0.85

 Reading

0.69

 planting

0.66

ħĭ

0.66

reader

0.66

FTP

0.65

 viewing

0.64

 readable

0.64

reading

0.63

Activations Density 0.000%

No Known Activations

This feature has no known activations.