---
license: gpl-3.0
license_name: gpl-3.0-or-later
license_link: https://github.com/LeelaChessZero/lc0/blob/01cc532648d9391e43bc006ca8ead7d966b9cc1b/README.md?plain=1#L184-L211
---

# Lc0 BT4-1024x15x32h-swa-6147500

- `config.yaml`: Originally `BT4-init1.yaml`, obtained from Discord.
- The model was originally downloaded from: https://storage.lczero.org/files/networks-contrib/big-transformers/BT4-1024x15x32h-swa-6147500.pb.gz

## Use with ChessNets

```python
from chessnets.pipelines.lc0 import Lc0Pipeline

pipeline = Lc0Pipeline.from_pretrained("shermansiu/lc0_BT4-1024x15x32h-swa-6147500")
predictions = pipeline("1. e4 e5")
```

## Checkpoint artifacts

- `config.json`: canonical, fully materialized ChessNets architecture config
- `model.safetensors`: decoded float32 PyTorch tensors
- `provenance.json`: source and conversion hashes
- `original/model.pb.gz`: original Lc0 protobuf checkpoint
- `training/original_lczero_training_config.yaml`: original training config

The learned protobuf weight layers used `LINEAR16` encoding. Conversion decoded
those values into float32 tensors; it cannot restore precision already lost in
the source encoding.

## Intended use and limitations

This is an Lc0 policy/value network. Its playing behavior and strength depend on the source network and the inference/search settings.

Review legal moves and probabilities in context. Outputs are not guarantees of
playing strength, fairness, or suitability for high-stakes decisions.

## Citation and attribution

Credit the source network identified above and the
[Leela Chess Zero project](https://lczero.org/). This community checkpoint does
not have a model-specific paper citation until the official project supplies one.