| ---
|
| license: apache-2.0
|
| tags:
|
| - co-attention
|
| - gelu
|
| - lamb
|
| - layernorm
|
| - linear-warmup
|
| - mocov3
|
| - nano
|
| - retrieval
|
| - standard
|
| - trunc-normal
|
| ---
|
|
|
| # finetune.py
|
|
|
| ## Model Overview
|
|
|
| A **nano**-scale implementation of the **mocov3** architecture, built for **retrieval** tasks.
|
|
|
| ## Architecture
|
|
|
| - **Architecture**: mocov3
|
| - **Scale**: nano
|
| - **Attention**: standard
|
| - **Fusion strategy**: co attention
|
| - **Task head**: retrieval
|
| - **Activation**: gelu
|
| - **Normalization**: layernorm
|
| - **Initialization**: trunc normal
|
|
|
| ## Training
|
|
|
| - **Optimizer**: lamb
|
| - **LR scheduler**: linear warmup
|
|
|
| ## Files
|
|
|
| - `finetune.py` — main artifact of this repository
|
|
|
| ## License
|
|
|
| See the license field above.
|
|
|
