INDEX

Explanations

information related to architecture and design projects

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_fs1536-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_1536

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

1,536

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

vag

-0.84

amen

-0.83

ighed

-0.81

ayed

-0.80

giving

-0.79

uggest

-0.79

subject

-0.79

rel

-0.79

take

-0.78

ogi

-0.78

POSITIVE LOGITS

 Wire

0.83

 Pulse

0.80

Gas

0.79

 Bout

0.79

Gul

0.77

Mud

0.77

 Rapids

0.76

 Butt

0.76

 Cout

0.76

 Lanka

0.75

Activations Density 2.271%

information related to architecture and design projects

No Comments

No Known Activations