You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
# Model Card for ThinkOnward's Geophysical Foundation Model
4
+
Welcome to the Geophysical Foundation Model (GFM) repository! This repository is designed to help you explore, understand, and use the Geophysical Foundation Model.
13
5
14
-
This model has been pushed to the Hub using the [PytorchModelHubMixin](https://huggingface.co/docs/huggingface_hub/package_reference/mixins#huggingface_hub.PyTorchModelHubMixin) integration:
6
+
## Table of Contents
7
+
1.[How to Fork and Clone the Repository](#how-to-fork-and-clone-the-repository)
8
+
2.[Getting Started](#getting-started)
9
+
3.[Running the Model](#running-the-model)
10
+
4.[Contributing](#contributing)
11
+
5.[Model Details](#model-details)
12
+
6.[Uses](#uses)
13
+
7.[Training Details](#training-details)
14
+
8.[Evaluation](#evaluation)
15
+
9.[Citations](#citations)
16
+
10.[Contact](#model-card-contact)
17
+
18
+
## How to Fork and Clone the Repository
19
+
20
+
If you want to work on this repository or make changes follow the instructions provided below.
21
+
22
+
### Step 1: Fork the Repository
23
+
To fork the repository to your GitHub account:
24
+
25
+
1. Go to the [Geophysical Foundation Model repository page](https://github.com/thinkonward/geophysical-foundation-model).
26
+
2. Click the "Fork" button located at the top right corner of the repository.
27
+
28
+
### Step 2: Clone the Forked Repository
29
+
Once you have forked the repository, you can clone it to your local machine.
30
+
31
+
1. Open a terminal or command prompt.
32
+
2. Navigate to the directory where you want to store the cloned repository.
33
+
3. Run the following command to clone the forked repository (remember to change yourusername to your actual GitHub username):
### Step 3: Change into the newly cloned directory:
40
+
41
+
```bash
42
+
cd geophysical-foundation-model
43
+
```
44
+
45
+
### Step 4: Create a new branch
46
+
47
+
Before making changes to the repository, it’s a good practice to create a new branch for your work. This keeps your main codebase clean and organized.
48
+
49
+
1. Create a new branch using the following command. Replace `feature-my-feature` with your desired branch name:
50
+
51
+
```bash
52
+
git checkout -b feature-my-feature
53
+
```
54
+
55
+
2. Now you are working on a separate branch and can make changes without affecting the main codebase.
56
+
57
+
## Getting Started
58
+
59
+
Now that you have forked and cloned the repository to your local machine and have created a new branch, you can start exploring the contents.
60
+
61
+
### Directory Structure
62
+
The project is organized as follows:
63
+
-`GFM/`: Contains model architecture.
64
+
-`src_imgs/`: images used in the repository.
65
+
-`Tutorial/`: A Jupyter notebook to get you started with the model.
66
+
-`README.md`: This file.
67
+
-`LICENSE`: The open source license file.
68
+
-`requirements.txt`: Dependencies for the model.
69
+
70
+
## Running the Model
71
+
72
+
1. Ensure you have Python (>3.10) installed on your system.
73
+
2. Start from the `geophysical-foundation-model` directory in the terminal.
74
+
3. Create a new virtual environment
75
+
```bash
76
+
conda create -n geophysics python=3.10
77
+
```
78
+
3. After creation, activate the newly created environment:
79
+
80
+
```bash
81
+
conda activate geophysics
82
+
```
83
+
4. Install the required dependencies by running:
84
+
85
+
```bash
86
+
pip install -r requirements.txt
87
+
```
88
+
5. Start up a Jupyterlab notebook and get started with the tutorial in the `Tutorial` directory. To start a notebook server from the terminal run
89
+
90
+
```bash
91
+
jupyter lab
92
+
```
93
+
94
+
And open `Tutorial/GFM_webinar.ipynb` to learn more about data processing.
95
+
96
+
97
+
6. For advanced users, you can download the model weights from HuggingFace and load them into an instance of the model in Python after you have requested access on HuggingFace.
# assuming you are in the `geophysical-foundation-model` directory
104
+
fromGFMimport ElasticViTMAE
105
+
model = ElasticViTMAE.ElasticViTMAE()
106
+
107
+
```
108
+
109
+
## Contributing
110
+
111
+
We welcome contributions from anyone interested in improving this project! To contribute to the model use the following steps:
112
+
113
+
1. Fork the repository.
114
+
2. Create a new branch for your changes:
115
+
116
+
```bash
117
+
git checkout -b feature/my-feature
118
+
```
119
+
120
+
3. Make your changes and commit them:
121
+
122
+
```bash
123
+
git add .
124
+
git commit -m "Add my feature"
125
+
```
126
+
127
+
4. Push your changes to your forked repository:
128
+
129
+
```bash
130
+
git push origin feature/my-feature
131
+
```
132
+
133
+
5. Create a pull request from your forked repository back to the original repository.
134
+
135
+
Thank you for contributing to the Geophysical Foundation Model
15
136
16
-
This is a model based on [Meta's ViTMAE model](https://huggingface.co/facebook/vit-mae-base), with some modifications to the masking technique. The Geophyiscal Foundation Model, or GFM for short, uses the ViT architecture with masking on traces in 2D seismic images, rather than patches.
17
137
18
138
## Model Details
19
139
@@ -23,7 +143,7 @@ ThinkOnward's Geophysical Foundation Model is a pre-trained a Vision Transformer
23
143
of seismic interpolation. We use 50 3D seismic volumes from the Patch the Planet Challenge, hosted by ThinkOnward as our benchmark hold-out dataset. **Using a Structural Similarity Index Metric (SSIM) to
24
144
compare results we document the Geophysical Foundation Model is 2-3 times better than Shang et al. (2023), and similar to Lasscock et al. (2024).**
25
145
26
-
-**Developed by:** Ognjen Tanovic and Mike McIntire of ThinkOnward (Shell Portfolio Company)
146
+
-**Developed by:** Ognjen Tanovic and Mike McIntire of ThinkOnward
27
147
-**Model type:** MAE
28
148
-**License:** Apache 2.0
29
149
-**Based on:** facebook/vit-mae-base
@@ -58,26 +178,6 @@ Downstream tasks include:
58
178
59
179
The backbone of this model was trained using 3D seismic data from the Patch the Planet Challenge hosted by ThinkOnward. Use of this model on anything outside of seismic data, or similar technologies would be out-of-scope and likely have poor performance.
0 commit comments