Instructions to use KaedeTai/dflash2-mlx-quantized-draft with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- MLX
How to use KaedeTai/dflash2-mlx-quantized-draft with MLX:
# Download the model from the Hub pip install huggingface_hub[hf_xet] huggingface-cli download --local-dir dflash2-mlx-quantized-draft KaedeTai/dflash2-mlx-quantized-draft
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
- Atomic Chat
fix: M5 Max is 128 GB, not 137 GB (byte-count/1e9 conversion error)
Browse files
README.md
CHANGED
|
@@ -30,7 +30,7 @@ Plus the measurements that motivated both.
|
|
| 30 |
|
| 31 |
## Headline: quantizing the *drafter* is close to free throughput
|
| 32 |
|
| 33 |
-
Apple **M5 Max**,
|
| 34 |
`tozp/Qwen3.8-27B-OBLITERATED-V2-mlx-4bit`. Greedy, 300 tokens, K=4, interleaved A/B with a
|
| 35 |
90 s cooldown before each measurement, two rounds.
|
| 36 |
|
|
|
|
| 30 |
|
| 31 |
## Headline: quantizing the *drafter* is close to free throughput
|
| 32 |
|
| 33 |
+
Apple **M5 Max**, 128 GB, macOS 26.4.1, MLX 0.32.0 / mlx-lm 0.31.3. Target:
|
| 34 |
`tozp/Qwen3.8-27B-OBLITERATED-V2-mlx-4bit`. Greedy, 300 tokens, K=4, interleaved A/B with a
|
| 35 |
90 s cooldown before each measurement, two rounds.
|
| 36 |
|