28 lines
1.1 KiB
Markdown
28 lines
1.1 KiB
Markdown
---
|
|
pipeline_tag: text-generation
|
|
base_model:
|
|
- LiquidAI/LFM2-24B-A2B
|
|
---
|
|
These are **MXFP4** quantizations of the model [LiquidAI / LFM2-24B-A2B](https://huggingface.co/LiquidAI/LFM2-24B-A2B)
|
|
|
|
## Quick Start
|
|
1. Download the latest release of **llama.cpp**.
|
|
2. Download your preferred model variant from below.
|
|
|
|
## Which version should I choose?
|
|
All variants use **MXFP4** for the MoE (Mixture of Experts) weights to keep the model efficient. The difference lies in how the remaining tensors are handled:
|
|
|
|
| Variant | Quality | Performance | Size | Recommendation |
|
|
| :--- | :--- | :--- | ---: | :--- |
|
|
| **BF16** | ⭐⭐⭐ | Variable* | 12.37GiB | Best for maximum accuracy; original unquantized weights. |
|
|
| **F16** | ⭐⭐ | Fast | 13.03GiB | Great alternative if BF16 is slow on your hardware. |
|
|
| **Q8** | ⭐ | Fastest | 13.03GiB | Balanced performance and memory usage. |
|
|
|
|
**Note:** On some older architectures, BF16 may be slower than F16.
|
|
Check that your GPU supports native BF16
|
|
|
|
Recommended parameters from LiquidAI:
|
|
- temperature 0.1
|
|
- top-k 50
|
|
- repetition_penalty 1.05
|