4-bit Open on Hugging Face
Parameters
73B
Weights (est.)
≈ 41.2 GB
Mac memory (est.)
64 GB+
Size is estimated from the parameter count and quantisation. Memory allows a fifth extra for context plus 2 GB for the system; long contexts need more.
Run it
With mlx-vlm on an Apple silicon Mac:
pip install mlx-vlmpython -m mlx_vlm.generate --model mlx-community/dolphin-vision-72b-4bit --image photo.jpg --prompt "Describe this image"
Details
- Base model
- Qwen/Qwen2-72B (finetune)
- Vendor
- Qwen
- Task
- Vision + text
- Quantisation
- 4-bit
- Licence
- tongyi-qianwen
- Downloads
- 921 in the last 30 days · 14K all time · 8 likes
- Published
- 2024-07-04 · updated 2024-07-04 (2y ago)
- Tags
- llava-qwen2, text-generation, axolotl, custom_code
Other MLX builds of Qwen/Qwen2-72B
| Build | Quantisation | Weights | Mac memory | Downloads |
|---|---|---|---|---|
| dolphin-2.9.2-qwen2-72b-2bit | 2-bit | ≈ 6.0 GB | 16 GB+ | 30 |
| dolphin-2.9.2-qwen2-72b-4bit | 4-bit | ≈ 9.7 GB | 16 GB+ | 37 |
| dolphin-2.9.2-qwen2-72b-8bit | 8-bit | ≈ 23.9 GB | 32 GB+ | 52 |
| dolphin-vision-72b-4bit (this one) | 4-bit | ≈ 41.2 GB | 64 GB+ | 921 |