Upload folder using huggingface_hub
Browse files- README.md +96 -0
- data/sample_train.h5 +3 -0
- upload.py +7 -0
- weights/experimental/ddpm.ckpt +3 -0
- weights/experimental/fp.ckpt +3 -0
- weights/experimental/vae.ckpt +3 -0
README.md
ADDED
@@ -0,0 +1,96 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
|
2 |
+
```markdown
|
3 |
+
# microgen3D
|
4 |
+
|
5 |
+
[](https://github.com/baskargroup/MicroGen3D)
|
6 |
+
|
7 |
+
## Dataset Summary
|
8 |
+
|
9 |
+
**microgen3D** is a dataset of 3D voxelized microstructures designed for training, evaluation, and benchmarking of generative models—especially Conditional Latent Diffusion Models (LDMs). It includes both synthetic (Cahn-Hilliard) and experimental microstructures with multiple phases (2 to 3). The voxel grids range from `64³` up to `128×128×64`.
|
10 |
+
|
11 |
+
The dataset consists of three microstructure types:
|
12 |
+
- **Experimental microstructures**
|
13 |
+
- **2-phase Cahn-Hilliard microstructures**
|
14 |
+
- **3-phase Cahn-Hilliard microstructures**
|
15 |
+
|
16 |
+
The two Cahn-Hilliard datasets are thresholded versions of the same simulation source. For each dataset type, we also provide pretrained generative model weights, comprising:
|
17 |
+
- `vae.ckpt` – Variational Autoencoder
|
18 |
+
- `fp.ckpt` – Feature Predictor
|
19 |
+
- `ddpm.ckpt` – Denoising Diffusion Probabilistic Model
|
20 |
+
|
21 |
+
---
|
22 |
+
|
23 |
+
## 📁 Repository Structure
|
24 |
+
|
25 |
+
```
|
26 |
+
microgen3D/
|
27 |
+
├── data/
|
28 |
+
│ └── sample_data.h5 # Experimental or synthetic HDF5 microstructure file
|
29 |
+
├── models/
|
30 |
+
│ └── weights/
|
31 |
+
│ ├── experimental/
|
32 |
+
│ │ ├── vae.ckpt
|
33 |
+
│ │ ├── fp.ckpt
|
34 |
+
│ │ └── ddpm.ckpt
|
35 |
+
│ ├── two_phase/
|
36 |
+
│ └── three_phase/
|
37 |
+
└── ...
|
38 |
+
```
|
39 |
+
|
40 |
+
---
|
41 |
+
|
42 |
+
## 🚀 Quick Start
|
43 |
+
|
44 |
+
### 🔧 Setup Instructions
|
45 |
+
|
46 |
+
```bash
|
47 |
+
# 1. Clone the repo
|
48 |
+
git clone https://github.com/baskargroup/MicroGen3D.git
|
49 |
+
cd MicroGen3D
|
50 |
+
|
51 |
+
# 2. Set up environment
|
52 |
+
python -m venv venv
|
53 |
+
source venv/bin/activate # On Windows use: venv\Scripts\activate
|
54 |
+
|
55 |
+
# 3. Install dependencies
|
56 |
+
pip install -r requirements.txt
|
57 |
+
|
58 |
+
# 4. Download dataset and weights (Hugging Face)
|
59 |
+
# Make sure HF CLI is installed and you're logged in: `huggingface-cli login`
|
60 |
+
```
|
61 |
+
|
62 |
+
```python
|
63 |
+
from huggingface_hub import hf_hub_download
|
64 |
+
|
65 |
+
# Download sample data
|
66 |
+
hf_hub_download(repo_id="BGLab/microgen3D", filename="sample_data.h5", repo_type="dataset", local_dir="data")
|
67 |
+
|
68 |
+
# Download model weights
|
69 |
+
hf_hub_download(repo_id="BGLab/microgen3D", filename="vae.ckpt", local_dir="models/weights/experimental")
|
70 |
+
hf_hub_download(repo_id="BGLab/microgen3D", filename="fp.ckpt", local_dir="models/weights/experimental")
|
71 |
+
hf_hub_download(repo_id="BGLab/microgen3D", filename="ddpm.ckpt", local_dir="models/weights/experimental")
|
72 |
+
```
|
73 |
+
|
74 |
+
---
|
75 |
+
|
76 |
+
## 📜 Citation
|
77 |
+
|
78 |
+
If you use this dataset or models, please cite:
|
79 |
+
|
80 |
+
```
|
81 |
+
@article{baishnab2025microgen3d,
|
82 |
+
title={3D Multiphase Heterogeneous Microstructure Generation Using Conditional Latent Diffusion Models},
|
83 |
+
author={Baishnab, Nirmal and Herron, Ethan and Balu, Aditya and Sarkar, Soumik and Krishnamurthy, Adarsh and Ganapathysubramanian, Baskar},
|
84 |
+
journal={arXiv preprint arXiv:2503.10711},
|
85 |
+
year={2025}
|
86 |
+
}
|
87 |
+
```
|
88 |
+
|
89 |
+
---
|
90 |
+
|
91 |
+
## ⚖️ License
|
92 |
+
|
93 |
+
This project is licensed under the **MIT License**.
|
94 |
+
|
95 |
+
---
|
96 |
+
|
data/sample_train.h5
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:83e32dae94dc93c54b3fbafbcccde90406b109739c3ca68f2d62ce67c7a1f11a
|
3 |
+
size 419821464
|
upload.py
ADDED
@@ -0,0 +1,7 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from huggingface_hub import upload_folder
|
2 |
+
|
3 |
+
upload_folder(
|
4 |
+
repo_id='BGLab/microgen3D',
|
5 |
+
repo_type="dataset",
|
6 |
+
folder_path=".",
|
7 |
+
)
|
weights/experimental/ddpm.ckpt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:af891210dd9adc62329e678d64807cb80c9bf17761d49b9ad1024774e2d791f4
|
3 |
+
size 1809545183
|
weights/experimental/fp.ckpt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:14bb6a355d1b0c6f2ca8128856cbb09d5438dfcf944c4bfda4b20109e6e09884
|
3 |
+
size 1985367
|
weights/experimental/vae.ckpt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:0cd90c3cc66bd9e9633ba3123388d3675aca1401c8e6ed5cba6b1cd02faedad1
|
3 |
+
size 563094509
|