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

emetery

-0.74

ascript

-0.70

auga

-0.70

agy

-0.69

adena

-0.66

 awfully

-0.63

 millenn

-0.62

arta

-0.61

ogi

-0.61

zona

-0.60

POSITIVE LOGITS

è£ħ

0.78

Pwr

0.72

LECT

0.71

burning

0.69

ãĥĩãĤ£

0.69

theless

0.67

 "$:/

0.66

²¾

0.66

mpeg

0.66

kj

0.65

Activations Density 0.000%

No Known Activations

This feature has no known activations.