Upload folder using huggingface_hub
Browse files- .gitignore +1 -0
- README.md +68 -0
.gitignore
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
.DS_Store
|
README.md
ADDED
@@ -0,0 +1,68 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# GH200 SEMamba Docker Image
|
2 |
+
|
3 |
+
This Docker image provides a pre-configured development environment for running [SEMamba](https://github.com/RoyChao19477/SEMamba), a mamba-based speech enhancement model, on NVIDIA GH200 hardware. It was created to save time and ensure reproducibility for ARM64/aarch64 CUDA 12.8 setups.
|
4 |
+
|
5 |
+
The image includes all required dependencies and configurations for Mamba-based speech enhancement research and other sequence modeling tasks.
|
6 |
+
|
7 |
+
---
|
8 |
+
|
9 |
+
## Contents
|
10 |
+
|
11 |
+
- **OS**: Ubuntu 24.04 (ARM64)
|
12 |
+
- **Python**: 3.12
|
13 |
+
- **CUDA**: 12.8
|
14 |
+
- **PyTorch**: 2.7
|
15 |
+
- **Mamba-SSM**: v2.2.4
|
16 |
+
- **FlashAttention**: v2.7.4.post1
|
17 |
+
*(Pytorch, Mamba-ssm, and flash-attn are all installed from [Jetson SBSA CUDA 12.8 index](https://pypi.jetson-ai-lab.dev/sbsa/cu128))*
|
18 |
+
- **Essential packages**: `ffmpeg`, `vim`, `htop`, `tmux`, `screen`, etc.
|
19 |
+
|
20 |
+
---
|
21 |
+
|
22 |
+
## Usage
|
23 |
+
|
24 |
+
### Load Docker Image
|
25 |
+
|
26 |
+
```bash
|
27 |
+
docker load < gh200_semamba_py312_pt27_cuda128.tar
|
28 |
+
```
|
29 |
+
|
30 |
+
### Run Container
|
31 |
+
|
32 |
+
```bash
|
33 |
+
docker run --gpus all -it -v $(pwd):/workspace gh200_semamba_py312_pt27_cuda128
|
34 |
+
```
|
35 |
+
|
36 |
+
This will mount your current directory into `/workspace` inside the container.
|
37 |
+
|
38 |
+
---
|
39 |
+
|
40 |
+
## Notes
|
41 |
+
|
42 |
+
- This image is intended **only for GH200 (ARM64) systems** with CUDA 12.8.
|
43 |
+
- Python packages are installed via a custom PyPI index for ARM64 provided by Jetson AI Lab.
|
44 |
+
- **Do not use this image on x86 systems**; it will not work.
|
45 |
+
- This environment was built to support projects like SEMamba that use modern state space models.
|
46 |
+
|
47 |
+
---
|
48 |
+
|
49 |
+
## License & Attribution
|
50 |
+
|
51 |
+
- This Docker image is shared for **non-commercial research purposes**.
|
52 |
+
- All third-party packages, including PyTorch, FlashAttention, and Mamba-SSM, retain their **original licenses**.
|
53 |
+
- PyTorch was installed from a community-provided index: https://pypi.jetson-ai-lab.dev
|
54 |
+
- Users are responsible for complying with the licenses of any included or downloaded components.
|
55 |
+
|
56 |
+
---
|
57 |
+
|
58 |
+
## Acknowledgments
|
59 |
+
|
60 |
+
Thanks to the developers of:
|
61 |
+
- Jetson AI Lab for maintaining ARM64-compatible PyTorch wheels.
|
62 |
+
|
63 |
+
---
|
64 |
+
|
65 |
+
## Maintainer
|
66 |
+
|
67 |
+
For any issues, feel free to open a discussion or contact me.
|
68 |
+
|