| --- |
| license: mit |
| tags: |
| - bitnet |
| - ternary |
| - trillim |
| - cpu-inference |
| base_model: microsoft/bitnet-b1.58-2B-4T-bf16 |
| --- |
| |
| # BitNet-TRNQ |
|
|
| Ternary-quantized version of [microsoft/bitnet-b1.58-2B-4T-bf16](https://huggingface.co/microsoft/bitnet-b1.58-2B-4T-bf16), packaged for the [Trillim DarkNet](https://huggingface.co/Trillim) inference engine. |
|
|
| This model runs entirely on CPU — no GPU required. |
|
|
| ## Model Details |
|
|
| | | | |
| |---|---| |
| | **Architecture** | BitNet (BitNetForCausalLM) | |
| | **Parameters** | ~2B | |
| | **Hidden size** | 2560 | |
| | **Layers** | 30 | |
| | **Attention heads** | 20 (5 KV heads) | |
| | **Context length** | 4096 | |
| | **Quantization** | Ternary ({-1, 0, 1}) | |
| | **Source model** | [microsoft/bitnet-b1.58-2B-4T-bf16](https://huggingface.co/microsoft/bitnet-b1.58-2B-4T-bf16) | |
| | **License** | MIT | |
|
|
| ## Usage |
|
|
| ```bash |
| pip install trillim |
| trillim pull Trillim/BitNet-TRNQ |
| trillim serve Trillim/BitNet-TRNQ |
| ``` |
|
|
| This starts an OpenAI-compatible API server at `http://127.0.0.1:8000`. |
|
|
| For interactive CLI chat: |
|
|
| ```bash |
| trillim chat Trillim/BitNet-TRNQ |
| ``` |
|
|
| ## What's in this repo |
|
|
| | File | Description | |
| |---|---| |
| | `qmodel.tensors` | Ternary-quantized weights in Trillim format | |
| | `rope.cache` | Precomputed RoPE embeddings | |
| | `config.json` | Model configuration | |
| | `tokenizer.json` | Tokenizer | |
| | `tokenizer_config.json` | Tokenizer configuration | |
| | `trillim_config.json` | Trillim metadata | |
|
|
| ## License |
|
|
| This model is released under the [MIT License](https://opensource.org/licenses/MIT), following the license of the source model. |
|
|
|
|
