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

omo

-0.66

ime

-0.63

 breach

-0.62

ulhu

-0.61

ylum

-0.60

ir

-0.60

erva

-0.60

HIT

-0.59

hiba

-0.59

iss

-0.58

POSITIVE LOGITS

Loading

0.79

Reviewer

0.79

Redditor

0.75

arser

0.72

MSN

0.70

lishes

0.70

IENT

0.69

Frag

0.68

Reply

0.66

SCP

0.66

Activations Density 0.000%

No Known Activations

This feature has no known activations.