INDEX

Explanations

references to commonly accepted ideas or standards

oai_token-act-pair · gpt-3.5-turbo Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res_fs768-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_768

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

768

Data Type

torch.float32

Hook Point

blocks.8.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

asus

-0.78

kamp

-0.74

rogram

-0.68

romeda

-0.67

Drug

-0.66

haw

-0.66

zik

-0.65

mented

-0.64

acus

-0.63

atoon

-0.63

POSITIVE LOGITS

ised

1.13

ization

1.09

isation

1.01

izations

0.96

 wisdom

0.94

 fare

0.93

cy

0.91

 deviation

0.90

ized

0.89

 sized

0.89

Activations Density 3.986%

references to commonly accepted ideas or standards

No Comments

No Known Activations