sciml's Introduction

SC21 Tutorial on Scientific Machine Learning

In this repository, you can find all the example notebooks used for SC21 full-day tutorial: Scientific Machine Learning using HPC Servers on Cloud

Resources
Target Audience
Content Level
Prerequisites
Getting Started
Frequently Asked Questions

Resources

Target Audience

Practitioners who use numerical simulations of Partial Differential Equations (PDEs) in analysis, optimization, design and control of complex engineered systems

Content level

20% Beginner, 40 % Intermediate, 40% Advanced

Prerequisites

Partial Differential Equations, Numerical methods, Machine Learning, Deep Learning, High Performance Computing, Python programming, Jupyter

Getting Started

Please email RocketML for a user account on sciml.rocketml.net. We will send you email instructions on how to log in.
Log in to sciml.rocketml.net using the instructions received from RocketML

Go through the onboarding screens
You will see a list of tutorials that are ready to use
Group by Topic to see Beginner, Intermediate, Advanced tutorials

Select a tutorial and wait for Jupyter Compute to start

Run the tutorial notebook one cell at a time. If you are not familiar with Jupyter notebook please google for a relevant tutorial

FAQ

My tutorial screen is stuck at Please Wait Jupyter Compute is not started yet for more than 5 minutes. What do I do?

This can happen due to following reasons:

When you log in for the first time. Resources like persistent disk space and Azure blob storage are being created for you to store the tutorials and for you to create new notebooks. For first login, expect up to 10 minutes delay.
There is a delay in creating containers when new nodes are being added to Azure Kubernetes cluster. A new node creation and downloading Docker images can take up to 10 minutes. If it takes more than 15 minutes, please ping us on slack.
Similar to system reboot for Windows that fixes all the issues, we have two fixes 1) Browser refresh, 2) Log out/Log in to get new containers.

sciml's People

Contributors

Stargazers

Watchers

sciml's Issues

AttributeError: 'function' object has no attribute 'gradients'

When running the following peace of code from sciml/04_PINNs/02 2D Heat Equation.ipynb:

t1 = time.time()

losshistory, train_state = model.train(epochs=10000)
t2 = time.time()
print("training time:", (t2-t1))

The following error arrises. Please fix this.

Warning: epochs is deprecated and will be removed in a future version. Use iterations instead.
Training model...

---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
Cell In[8], line 3
      1 t1 = time.time()
----> 3 losshistory, train_state = model.train(epochs=10000)
      4 t2 = time.time()
      5 print("training time:", (t2-t1))

File [~/miniconda3/envs/physxml/lib/python3.8/site-packages/deepxde/utils/internal.py:22](https://file+.vscode-resource.vscode-cdn.net/home/n1ghtf4l1/Workspace/sciml-main/04_PINNs/~/miniconda3/envs/physxml/lib/python3.8/site-packages/deepxde/utils/internal.py:22), in timing..wrapper(*args, **kwargs)
     19 @wraps(f)
     20 def wrapper(*args, **kwargs):
     21     ts = timeit.default_timer()
---> 22     result = f(*args, **kwargs)
     23     te = timeit.default_timer()
     24     if config.rank == 0:

File [~/miniconda3/envs/physxml/lib/python3.8/site-packages/deepxde/model.py:628](https://file+.vscode-resource.vscode-cdn.net/home/n1ghtf4l1/Workspace/sciml-main/04_PINNs/~/miniconda3/envs/physxml/lib/python3.8/site-packages/deepxde/model.py:628), in Model.train(self, iterations, batch_size, display_every, disregard_previous_best, callbacks, model_restore_path, model_save_path, epochs)
    626 self.train_state.set_data_train(*self.data.train_next_batch(self.batch_size))
    627 self.train_state.set_data_test(*self.data.test())
--> 628 self._test()
    629 self.callbacks.on_train_begin()
    630 if optimizers.is_external_optimizer(self.opt_name):

File [~/miniconda3/envs/physxml/lib/python3.8/site-packages/deepxde/model.py:817](https://file+.vscode-resource.vscode-cdn.net/home/n1ghtf4l1/Workspace/sciml-main/04_PINNs/~/miniconda3/envs/physxml/lib/python3.8/site-packages/deepxde/model.py:817), in Model._test(self)
    812 def _test(self):
...
---> 67     du_X = tf.gradients(u, X)[0]
     68     du_x, du_y, du_t = du_X[:, 0:1], du_X[:, 1:2],du_X[:, 2:3]
     69     du_xx  = tf.gradients(du_x, X)[0][:, 0:1]

AttributeError: 'function' object has no attribute 'gradients'

System Information:
Python v3.8.16
OS Endeavor OS
DeepXDE v1.9.3

Recommend Projects