MLX Community · First MLX build 24 Aug 2024
Phi-3.5-MoE-instruct
MLX
Text mit
A 42B parameter text generation model, with 3 MLX builds, from 4-bit to bf16.
- Parameters
- 42B From Hugging Face
- Quantisation
- 4-bit Phi-3.5-MoE-instruct-4bit
- Memory needed
- 29 GB Estimate, unified memory
- Download
- 23.6 GB Weights, estimated
- Downloads
- 725 This build, last 30 days
- Task
- Text generation Text
- MLX builds
- 3 On mlx-community
- Devices
- Mac Apple silicon
Memory
Will it fit?Needs about 29 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/Phi-3.5-MoE-instruct-4bit --prompt "Hello"MLX builds
3 on mlx-community| Phi-3.5-MoE-instruct-4bit | 4-bit | 23.6 GB | 29 GB | 725 |
|---|---|---|---|---|
| Phi-3.5-MoE-instruct-8bit | 8-bit | 44.5 GB | 52 GB | 219 |
| Phi-3.5-MoE-instruct-bf16 | bf16 | 83.7 GB | 96 GB | 65 |
Sizes are estimated from Hugging Face’s parameter counts and the quantisation. Downloads are for the last 30 days.