terrabyte Installation

terrabyte

terrabyte is an innovative High Performance Data Analytics (HPDA) platform operated by the DLR and the Leibniz Supercomputing Center (LRZ), an institution of the Bavarian Academy of Sciences and Humanities. The platform provides researchers with efficient access to Earth Observation data, a powerful processing environment, and practical tools for data analysis.

Get an account

To request an account, you must either be

• a DLR employee
• part of one of the DLR cooperation chairs
• external DLR-partner for projects with DLR involvement
• part of a Bavarian research institution

Follow the instructions on the official terrabyte Documentation to get an account.

Start a Session

1. Login and navigate to My Interactive Sessions.

2. Choose Code Server for VS Code or Jupyther Notebook

3. Select gpu@hpda2_compute_gpu as cluster partition

4. Select the other properties as needed.

5. Launch the server.

SIPT Installation

1. Create a new Jupyter Notebook.

2. Install the SIPT library

   !pip install sipt[gpu] --index-url https://sipt:gldt-opt8vHHZkQ5kvTyDSetK@gitlab2.informatik.uni-wuerzburg.de/api/v4/projects/21379/packages/pypi/simple

3. If issues with CUDA occure, load the CUDA libraries manually

   Python

   import ctypes
   import os
   import glob
   
   cuda_lib_dir = "/usr/local/cuda/lib64"
   so_files = [f for f in glob.glob(os.path.join(cuda_lib_dir, "*.so")) if os.path.isfile(f)]
   failed_libs = {lib: None for lib in so_files}
   for attempt in range(4):
       for lib_path in list(failed_libs.keys()):
           try:
               ctypes.CDLL(lib_path)
               del failed_libs[lib_path]
           except Exception as e:
               failed_libs[lib_path] = str(e) 
               if attempt == 3: print(f"Failed to load {lib_path}: {str(e)}")
       if not failed_libs:
           print("\nAll libraries loaded successfully!")
           break