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

nesota

-0.67

 peninsula

-0.64

 opioid

-0.62

 opio

-0.61

 resignation

-0.60

 convoy

-0.59

adh

-0.59

 dispensary

-0.59

 exha

-0.59

 hosp

-0.58

POSITIVE LOGITS

oresc

0.81

angelo

0.81

LES

0.73

OHN

0.73

oret

0.72

olkien

0.71

IAS

0.70

anes

0.70

mus

0.70

.;

0.67

Activations Density 0.000%

No Known Activations

This feature has no known activations.