{"id":492,"date":"2018-10-02T17:05:58","date_gmt":"2018-10-02T16:05:58","guid":{"rendered":"http:\/\/ri.itservices.manchester.ac.uk\/csf3\/?page_id=492"},"modified":"2025-08-01T10:42:19","modified_gmt":"2025-08-01T09:42:19","slug":"keras","status":"publish","type":"page","link":"https:\/\/ri.itservices.manchester.ac.uk\/csf3\/software\/applications\/keras\/","title":{"rendered":"Keras"},"content":{"rendered":"

Overview<\/h2>\n

Keras<\/a> is a high-level neural networks API, written in Python that can run on top of TensorFlow<\/a> installed on the CSF.<\/p>\n

Restrictions on use<\/h2>\n

There are no restrictions on accessing this software on the CSF. The software is released under the MIT License<\/a> and all usage must adhere to that license.<\/p>\n

Set up procedure<\/h2>\n

Since version 2.4, Keras has been packaged within<\/a> Tensorflow as tensorflow.keras. To load Keras please load one<\/em> of the following modulefiles:<\/p>\n

# Backend is Tensorflow 2.4.0 (Python 3.7):\r\nmodule load apps\/binapps\/tensorflow\/2.4.0-37-cpu\r\nmodule load apps\/binapps\/tensorflow\/2.4.0-37-gpu<\/pre>\n
To access the older versions of this software you must first load one<\/em> of the following modulefiles:<\/p>\n
# Backend is Tensorflow 1.14.0 (Python 3.6) and the tf modulefile will be loaded for you:\r\nmodule load apps\/binapps\/keras\/2.2.4-tensorflow-gpu      # GPU version of tensorflow 1.14.0\r\nmodule load apps\/binapps\/keras\/2.2.4-tensorflow-cpu      # CPU version of tensorflow 1.14.0\r\n\r\n# Backend is Tensorflow 1.11.0 (Python 3.6) and the tf modulefile will be loaded for you:\r\nmodule load apps\/binapps\/keras\/2.2.2-tensorflow-gpu      # GPU version of tensorflow 1.11.0\r\nmodule load apps\/binapps\/keras\/2.2.2-tensorflow-cpu      # CPU version of tensorflow 1.11.0\r\n\r\n# Backend is Tensorflow but no<\/strong> tensorflow modulefile loaded. You must<\/strong>\r\n# load a tensorflow modulefile first<\/em>. Check the list of available versions<\/a> on CSF.\r\nmodule load apps\/binapps\/keras\/2.2.4\r\nmodule load apps\/binapps\/keras\/2.2.2\r\n<\/pre>\n
The keras modulefile will automatically load the tensorflow and anaconda python modulefiles for you unless otherwise indicated.<\/p>\n
Running the application<\/h2>\n
Please do not run Keras (via python) on the login node. Jobs should be submitted to the compute nodes via batch.<\/p>\n
Interactive use on a Backend Node<\/h3>\n
To request an interactive session on a backend compute node run:<\/p>\n
\r\n# CPU-only jobs (a one-hour interactive session)\r\nsrun -p interactive -t 0-1 --pty bash\r\n  # Wait until you are logged in to a compute-node, then:\r\n  module load apps\/binapps\/keras\/2.2.4-tensorflow-cpu\r\n\r\n  python\r\n  >>> [type your python code interactively]\r\n\r\n  # or run a python script you are developing, for example\r\n  python myscript.py<\/em>\r\n\r\n# GPU jobs (you may use up to 4 GPUs depending on your granted access to the GPUs)\r\nsrun -p gpuV -G 1 -t 0-1 --pty bash<\/strong>\r\n  # Wait until you are logged in to a gpu-node, then:\r\n  module load apps\/binapps\/keras\/2.2.4-tensorflow-gpu\r\n\r\n  python\r\n  >>> [type your python code interactively]\r\n\r\n  # or run a python script you are developing, for example\r\n  python myscript.py<\/em>\r\n<\/pre>\n
An example Keras session is given below.<\/p>\n
If there are no free interactive<\/em> resources the srun<\/code> command will ask you to try again later. Please do not run Keras (python) on the login node. Any jobs running there will be killed without warning.<\/p>\n
Example Script (for CPU or GPU)<\/h3>\n
The following skeleton script can be used in an interactive session or in a batch job (from a jobscript). It ensures tensorflow does not use more cores than you have requested in your jobscript. If run on a GPU-node, it will automatically use the GPU.<\/p>\n
# Some of the following code has been taken from the Keras examples at:\r\n# https:\/\/keras.io\/getting-started\/sequential-model-guide\/\r\n\r\nfrom keras import backend as K\r\nfrom keras.models import Sequential\r\nfrom keras.layers import Dense, Dropout\r\nimport numpy as np\r\nimport tensorflow as tf\r\nimport os\r\n\r\n# Get number of cores reserved by the batch system\r\n# ($SLURM_NTASKS is set by the batch system, or use 1 otherwise)\r\nNUMCORES=int(os.getenv(\"SLURM_NTASKS\",1))\r\nprint(\"Using\", NUMCORES, \"core(s)\" )\r\n\r\n# Create TF session using correct number of cores.\r\n# NOTE:\r\n#   If you are using the GPU version, this will also automatically use\r\n#   the GPU and set up the Keras CUDA libraries. Please see the CSF3\r\n#   tensorflow page for other GPU-specific tf.ConfigProto() entries.\r\n\r\nsess = tf.Session(config=tf.ConfigProto(inter_op_parallelism_threads=NUMCORES,\r\n   allow_soft_placement=True, device_count = {'CPU': NUMCORES}))\r\n\r\n# Set the Keras TF session\r\nK.set_session(sess)\r\n\r\n# Replace the rest of the script with your own code<\/strong>\r\n\r\n# Generate dummy data\r\nx_train = np.random.random((1000, 20))\r\ny_train = np.random.randint(2, size=(1000, 1))\r\nx_test = np.random.random((100, 20))\r\ny_test = np.random.randint(2, size=(100, 1))\r\n\r\n# MLP for binary classification example\r\nmodel = Sequential()\r\nmodel.add(Dense(64, input_dim=20, activation='relu'))\r\nmodel.add(Dropout(0.5))\r\nmodel.add(Dense(64, activation='relu'))\r\nmodel.add(Dropout(0.5))\r\nmodel.add(Dense(1, activation='sigmoid'))\r\n\r\nmodel.compile(loss='binary_crossentropy', optimizer='rmsprop', metrics=['accuracy'])\r\nmodel.fit(x_train, y_train, epochs=20, batch_size=128)\r\nscore = model.evaluate(x_test, y_test, batch_size=128)\r\nprint(\"Score\", score)\r\n<\/pre>\n
Serial CPU with a GPU batch job submission<\/h3>\n
Make sure you have the modulefile loaded then create a batch submission script, for example:<\/p>\n
\r\n#!\/bin\/bash --login\r\n#SBATCH -p gpuV       # Run on the v100 GPU nodes\r\n#SBATCH -G 1          # Use one GPU\r\n#SBATCH -n 1          # (or --ntasks=) Number of CPU cores\r\n#SBATCH -t 2-0        # Wallclock time limit (2-0 is two day, max permitted is 4 days.)\r\n\r\n# We now recommend loading the modulefile in the jobscript (change version as required)\r\nmodule load apps\/binapps\/keras\/2.2.4-tensorflow-cpu     # or use the -gpu version\r\n\r\n# $SLURM_NTASKS is automatically set to the number of cores requested on the pe line\r\n# and can be read by your python code.\r\nexport OMP_NUM_THREADS=$SLURM_NTASKS\r\npython my-script.py\r\n<\/pre>\n
Submit the jobscript using:<\/p>\n
sbatch scriptname<\/em><\/pre>\n
where scriptname<\/em> is the name of your jobscript.<\/p>\n
Parallel CPU with a GPU batch job submission<\/h3>\n
Ensure you have loaded the correct modulefile and then create a jobscript similar to the following:<\/p>\n
#!\/bin\/bash --login\r\n#SBATCH -p gpuV       # Run on the v100 GPU nodes\r\n#SBATCH -G 1          # Use one GPU\r\n#SBATCH -n 8          # (or --ntasks=) Number of CPU cores (can be up to 8 per v100 GPU)\r\n#SBATCH -t 2-0        # Wallclock time limit (2-0 is two day, max permitted is 4 days.)\r\n\r\n# We now recommend loading the modulefile in the jobscript (change version as required)\r\nmodule load apps\/binapps\/keras\/2.2.4-tensorflow-cpu     # or use the -gpu version\r\n\r\n# $SLURM_NTASKS is automatically set to the number of cores requested on the pe line\r\n# and can be read by your python code.\r\nexport OMP_NUM_THREADS=$SLURM_NTASKS\r\npython my-script.py\r\n<\/pre>\n
The above my-script.py<\/code> example will get the number of cores to use from the $SLURM_NTASKS<\/code> environment variable.<\/p>\n
Submit your jobscript using<\/p>\n
sbatch jobscript<\/em><\/pre>\n
where jobscript<\/em><\/code> is the name of your jobscript.<\/p>\n
Further info<\/h2>\n