--- base_model: MiniMaxAI/MiniMax-M2 base_model_relation: quantized quantized_by: ArtusDev pipeline_tag: text-generation license: mit tags: - exl3 ---
EXL3 quants of MiniMaxAI/MiniMax-M2 using exllamav3 for quantization.
⚠️ Requires ExLlamaV3 v0.0.12. ⚠️
| Quant | BPW | Head Bits | Size (GB) |
|---|---|---|---|
| 1.0_H6 | 1.0 | 6 | 30.8 |
| 1.09_H6 (optimized) | 1.09 | 6 | 32.8 |
| 1.49_H6 (optimized) | 1.49 | 6 | 44.0 |
| 2.0_H6 | 2.0 | 6 | 59.2 |
| 2.39_H6 (optimized) | 2.39 | 6 | 59.2 |
| 2.5_H6 | 2.5 | 6 | 73.40 |
| 2.76_H7 (optimized) | 2.76 | 7 | 80.27 |
| 3.0_H6 | 3.0 | 6 | 87.63 |
| 3.22_H6 (optimized) | 3.22 | 6 | 93.48 |
| 3.5_H6 | 3.5 | 6 | 101.83 |
| 3.68_H7 (optimized) | 3.68 | 7 | 106.54 |
| 4.0_H6 | 4.0 | 6 | 116.06 |
| 4.25_H6 | 4.25 | 6 | 123.16 |
| 5.0_H6 | 5.0 | 6 | 144.1 |
| 6.0_H6 | 6.0 | 6 | 173.22 |
| 8.0_H8 | 8.0 | 8 | 229.92 |
You can download quants by targeting specific size using the Hugging Face CLI.
pip install -U "huggingface_hub[cli]"
2. Download a specific quant:
huggingface-cli download ArtusDev/MiniMaxAI_MiniMax-M2-EXL3 --revision "5.0bpw_H6" --local-dir ./
EXL3 quants can be run with any inference client that supports EXL3, such as TabbyAPI. Refer to documentation for set up instructions.
Made possible with cloud compute from lium.io