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

 unloaded

-0.71

 extrap

-0.68

 looting

-0.68

 tampering

-0.68

bye

-0.67

 smugg

-0.64

 accumulated

-0.64

 descend

-0.64

 looted

-0.63

 Hebdo

-0.60

POSITIVE LOGITS

groups

0.76

ITNESS

0.75

raph

0.75

ombat

0.75

Raz

0.72

DERR

0.72

rament

0.72

RD

0.69

 Aires

0.69

fecture

0.68

Activations Density 0.000%

No Known Activations

This feature has no known activations.