File size: 1,558 Bytes
cb9c894 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 |
---
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.
|