readme update

examples/FNLGMD/README.md

@@ -13,7 +13,7 @@ model_type: 'jtnn-fnl'
 smiles_input_file: './zinc_smiles.txt'
 output_directory: './outputs/'
 scorer_type: 'LogPTestCase'
-num_epochs: 20 
+num_epochs: 2 
 
 #Selection/Optimization params
 optimizer_type: 'geneticoptimizer'
@@ -26,8 +26,8 @@ optima_type: "maxima"
 selection_type: "tournament"
 elite_perc: 0
 
-initial_pop_size: 100
-max_population: 500
+initial_pop_size: 10
+max_population: 10
 
 #FNL JTNN specific
 vocab_path: './all_vocab.txt'
@@ -38,6 +38,42 @@ Since all the input files will be copied to the cluster vm, they should not resi
 so make sure to copy the zinc_smiles.txt file from the other example directory. The output directory is set to
 the default AML output directory: ./outputs/.
 
-# running with a cutom environment
+The next step is to create the runscript.sh. This is tweaked a little but from the run_gmd.sh which can be found in the root of the repository.
+```
+export PATH=/opt/conda/envs/env/bin:$PATH
+source activate gmd
+python /FNLGMD/source/main.py -config config.yaml
+```
+
+The content of the working directory should now look like this:
+```
+(azureml_py38) azureuser@humeilan1:~/cloudfiles/code/Users/humeilan/FNLGMD/examples/LogP_JTVAE$ ls -alh
+total 44M
+drwxrwxrwx 2 root root    0 Dec 18 19:19 .
+drwxrwxrwx 2 root root    0 Dec 18 19:19 ..
+-rwxrwxrwx 1 root root  251 Dec 18 19:19 README.md
+-rwxrwxrwx 1 root root  525 Dec 18 19:19 Slurm-Run-Loop
+-rwxrwxrwx 1 root root  11M Dec 18 19:19 all.txt
+-rwxrwxrwx 1 root root 9.5K Dec 18 19:19 all_vocab.txt
+-rwxrwxrwx 1 root root  471 Dec 19 07:14 config.yaml
+-rwxrwxrwx 1 root root 2.0K Dec 18 19:19 example.out
+-rwxrwxrwx 1 root root  23M Dec 18 19:19 model.epoch-35
+-rwxrwxrwx 1 root root  112 Dec 19 07:14 runscript.sh
+-rwxrwxrwx 1 root root  11M Dec 18 20:30 zinc_smiles.txt
+```
+
+# running with a custom environment
+
+The above example input file requires the latest version of the code to run, hence the need to build the docker container ourselves.
+This can easily be done based on the Dockerfile and both dependency files provided in the repo. Create a new Environment, set the name to fnlgmd
+and select "Create a new docker context". Add the 3 required files and finish. The build will take a few minutes.
+
+
+The job can now be submitted with:
+```
+$ sbatch -p f16s --environment="fnlgmd@latest" --datamover=simple ./runscript.sh
+Uploading LogP_JTVAE (45.82 MBs): 100%|████████████████████████████████████████████| 45816548/45816548 [00:00<00:00, 118097961.49it/s]
+tough_candle_ld3n0b2lxp
+```
 
-   
+This should be a very quick job and the result can be found in the jobs "Outputs and logs" tab: