INDEX

Explanations

technical terms and concepts related to software development and technology

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

atos

-0.67

 Occupations

-0.66

bowl

-0.66

HAM

-0.64

ornia

-0.63

 Travels

-0.63

 Purchase

-0.62

atted

-0.62

pat

-0.62

irable

-0.61

POSITIVE LOGITS

 nonetheless

2.00

 nevertheless

1.64

etheless

1.58

 still

0.97

 undeniably

0.97

 undeniable

0.92

 cautioned

0.90

 caveats

0.88

 awfully

0.88

 caution

0.87

Activations Density 7.026%

technical terms and concepts related to software development and technology

No Comments

No Known Activations