mikovalenko's picture
Upload 2 files
28f28f9 verified
|
Raw
History Blame Contribute Delete
811 Bytes
---
license: mit
tags:
- gated-fusion
- grouped-query
- kaiming
- lamb
- mish
- multitask
- nano
- polynomial
- scalenorm
- tiny-transformer
---
# model_301040191_tiny_transformer_nano.py
## Model Overview
A **nano**-scale implementation of the **tiny transformer** architecture, built for **multitask** tasks.
## Architecture
- **Architecture**: tiny transformer
- **Scale**: nano
- **Attention**: grouped query
- **Fusion strategy**: gated fusion
- **Task head**: multitask
- **Activation**: mish
- **Normalization**: scalenorm
- **Initialization**: kaiming
## Training
- **Optimizer**: lamb
- **LR scheduler**: polynomial
## Files
- `model_301040191_tiny_transformer_nano.py` — main artifact of this repository
## License
See the license field above.