BigCode · First MLX build 29 Apr 2024
starcoder2-15b
MLX
Text bigcode-openrail-m
A 16B parameter text generation model from BigCode, with 8 MLX builds, from 4-bit to fp32.
- Parameters
- 16B From Hugging Face
- Quantisation
- 4-bit bigcode-starcoder2-15b-4bit
- Memory needed
- 13 GB Estimate, unified memory
- Download
- 9.0 GB Weights, estimated
- Downloads
- 199 This build, last 30 days
- Task
- Text generation Text
- MLX builds
- 8 On mlx-community
- Devices
- Mac Apple silicon
Memory
Will it fit?Needs about 13 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/bigcode-starcoder2-15b-4bit --prompt "Hello"MLX builds
8 on mlx-community| bigcode-starcoder2-15b-4bit | 4-bit | 9.0 GB | 13 GB | 199 |
|---|---|---|---|---|
| starcoder2-15b-instruct-v0.1-4bit | 4-bit | 9.0 GB | 13 GB | 106 |
| bigcode-starcoder2-15b-6bit | 6-bit | 13.0 GB | 17 GB | 82 |
| bigcode-starcoder2-15b-8bit | 8-bit | 17.0 GB | 21 GB | 98 |
| starcoder2-15b-instruct-v0.1-8bit | 8-bit | 17.0 GB | 21 GB | 74 |
| starcoder2-15b-instruct-v0.1 | Unstated | 31.9 GB | 38 GB | 95 |
| bigcode-starcoder2-15b-fp16 | fp16 | 31.9 GB | 38 GB | 84 |
| bigcode-starcoder2-15b-fp32 | fp32 | 63.8 GB | 74 GB | 186 |
Sizes are estimated from Hugging Face’s parameter counts and the quantisation. Downloads are for the last 30 days.