Views
No views yet
| Property | Value |
|---|---|
| Parameters | 1.2B |
| Precision | 5-bit |
| Group Size | 64 |
| Context Length | 128K |
| Parameter | Value |
|---|---|
| temperature | 0.3 |
| min_p | 0.15 |
| repetition_penalty | 1.05 |
| max_tokens | 512 |
pip install mlx-lm1from mlx_lm import load, generate
2from mlx_lm.sample_utils import make_sampler, make_logits_processors
3
4model, tokenizer = load("LiquidAI/LFM2.5-1.2B-JP-5bit")
5
6prompt = "東京は日本の"
7
8sampler = make_sampler(temp=0.3, min_p=0.15)
9logits_processors = make_logits_processors(repetition_penalty=1.05)
10
11response = generate(
12 model,
13 tokenizer,
14 prompt=prompt,
15 max_tokens=512,
16 sampler=sampler,
17 logits_processors=logits_processors,
18 verbose=True,
19)