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.