---
license: apache-2.0
tags:
- adafactor
- approx-gelu
- cosine
- grouped-query
- multitask
- poolformer
- rmsnorm
- tensor-fusion
- xavier-uniform
- xlarge
---

# predict.py

## Model Overview

A **xlarge**-scale implementation of the **poolformer** architecture, built for **multitask** tasks.

## Architecture

- **Architecture**: poolformer
- **Scale**: xlarge
- **Attention**: grouped query
- **Fusion strategy**: tensor fusion
- **Task head**: multitask
- **Activation**: approx gelu
- **Normalization**: rmsnorm
- **Initialization**: xavier uniform

## Training

- **Optimizer**: adafactor
- **LR scheduler**: cosine

## Files

- `predict.py` — main artifact of this repository

## License

See the license field above.