Alibaba Qwen · First MLX build 18 Sept 2024
Qwen2.5-1.5B
MLX
Text apache-2.0
A 1.5B parameter text generation model from Alibaba Qwen, with 7 MLX builds, from 2-bit to bf16.
- Parameters
- 1.5B From Hugging Face
- Quantisation
- 4-bit Qwen2.5-1.5B-Instruct-4bit
- Memory needed
- 3 GB Estimate, unified memory
- Download
- 869 MB Weights, estimated
- Downloads
- 11K This build, last 30 days
- Task
- Text generation Text
- MLX builds
- 7 On mlx-community
- Devices
- Mac Apple silicon
Memory
Will it fit?Needs about 3 GB at 4-bit
8 GB
16 GB
24 GB
32 GB
64 GB
128 GB
Fits with headroomTight, close other appsWon’t fit
Estimated from the size of the weights, plus a fifth for context and working memory and 2 GB for macOS and your apps. Long contexts need more.
Run it
Terminal · mlx-lm$ pip install -U mlx-lm
$ mlx_lm.generate --model mlx-community/Qwen2.5-1.5B-Instruct-4bit --prompt "Hello"MLX builds
7 on mlx-community| Josiefied-Qwen2.5-1.5B-Instruct-abliterated-v1-2bit | 2-bit | 483 MB | 3 GB | 47 |
|---|---|---|---|---|
| Qwen2.5-1.5B-Instruct-4bit | 4-bit | 869 MB | 3 GB | 11K |
| Josiefied-Qwen2.5-1.5B-Instruct-abliterated-v1-4bit | 4-bit | 869 MB | 3 GB | 103 |
| Qwen2.5-1.5B-Instruct-8bit | 8-bit | 1.6 GB | 4 GB | 146 |
| Josiefied-Qwen2.5-1.5B-Instruct-abliterated-v1-8bit | 8-bit | 1.6 GB | 4 GB | 42 |
| Qwen2.5-1.5B-Instruct-bf16 | bf16 | 3.1 GB | 6 GB | 295 |
| Josiefied-Qwen2.5-1.5B-Instruct-abliterated-v1-f16 | Unstated | 3.1 GB | 6 GB | 43 |
Sizes are estimated from Hugging Face’s parameter counts and the quantisation. Downloads are for the last 30 days.