audio model · musicgen · Linux
Can I run MusicGen large on Nvidia DGX Spark (128GB)?
Yes. MusicGen large runs on Nvidia DGX Spark (128GB) at fp32 (~20 GB of ~112 GB usable).
Runs at fp32 using ~20 GB of ~112 GB usable.
- Peak memory
- ~20 GB
- Usable on device
- ~112 GB
- Device memory
- 128 GB
- Quant
- fp32
How to run it
Use AudioCraft or HF Transformers at fp32. It loads as a single model and runs on a GPU.
- Type
- Music generation
- Parameters
- 3.3B
- Peak memory
- ~20 GB at fp32
- License
- CC-BY-NC-4.0
- Memory
- 128 GB unified
- Usable for weights
- ~112 GB
- Power draw
- ~140 W
- Best runtime
- llama.cpp (CUDA) / Ollama (CUDA)
You could also run
Run MusicGen large on other hardware
FAQ
Can Nvidia DGX Spark (128GB) run MusicGen large?
Yes. MusicGen large runs on Nvidia DGX Spark (128GB) at fp32 (~20 GB of ~112 GB usable).
How much memory does MusicGen large need?
Nvidia DGX Spark (128GB) has room to spare. At fp32 the realistic peak is ~20 GB of memory.
What do I use to run MusicGen large locally?
MusicGen large runs in AudioCraft or HF Transformers. It needs a GPU.
Sources
-
docs.nvidia.com · 2 sources
-
github.com · 2 sources
-
huggingface.co · 1 source
-
nvidia.com · 1 source
-
nvidianews.nvidia.com · 1 source
-
ollama.com · 1 source
-
tomshardware.com · 1 source
VRAM figures are sourced peak-usage anchors at the noted quant; catalog updated 2026-11-01. See methodology.