Skip to content

Files

Latest commit

 

History

History
136 lines (136 loc) · 7.33 KB

TODO.md

File metadata and controls

136 lines (136 loc) · 7.33 KB

To do list

  • Add workflow badge to README
  • Update Auxillary models
  • Remove spacing between bottom of func comments and 1st line of func.
  • Change globals module to _globals in psp dir.
  • Append each jobs results to a the one csv file including model and GCP parameters.
  • Remove URL and filenames for datasets from globals.
  • Visualise GCP code pipeline
  • Create model version on AI Platform
  • Add API's required for GCP part
  • Add roles required on GCP.
  • Create front-end React App that receives input from the finished job and results for each job and visualises and returns them to a front-end web app thing.
  • Update notification func to update when job failed and parse reason.
  • Add Releases
  • Update function comments
  • In workflow, test code pipeline by running dummy model and checking resultant files etc.
  • Continue Hyperparameter tuning of model
  • Add https://drive.google.com/drive/folders/1404cRlQmMuYWPWp5KwDtA7BPMpl-vF-d to Data Section
  • Fix README's
  • Check Latest Travis Build
  • Add AUC() metric class to models
  • Change colour of box in boxplot
  • Fix Boxplots - what do they represent etc...
  • Model Tests
  • Tests for inputting data for prediction - fasta, txt, pdb tests, add data folder in tests folder
  • Add learning rate scheduler
  • Add labels to readme
  • Add CI Github workflows
  • Add CI Testing - https://docs.github.com/en/free-pro-team@latest/actions/guides/building-and-testing-python#introduction
  • Add AUC, FP and FN to output metrics
  • Coveralls - https://coveralls.io/
  • Review one-hot encoding process
  • Review neccisity of all_data variable
  • Reach out to ICML people and find out how they developed their data
  • H/w requirements in readme
  • Look into pytest
  • CodeCov - Code Coverage
  • Python Version Badge - https://shields.io/category/platform-support
  • Last Modified Badge - https://shields.io/category/activity
  • LinkedIn Badge
  • GCP Badge - https://img.shields.io/badge/Google_Cloud-4285F4?style=for-the-badge&logo=google-cloud&logoColor=white
  • Python Logo Badge
  • Visualise Keras model - https://www.machinecurve.com/index.php/2019/10/07/how-to-visualize-a-model-with-keras/
  • Re do model tests
  • Remove TensorBaord stuff from model and only keep in training file
  • Keras JSON Parser
  • Check variable and layer names for models
  • Remove GCP config script
  • Add Workflow tests for psp_gcp whereby gcloud sdk is installed and a few commands are attempted to see if it is working correctly etc
  • Remove show plots parameter #unnessary
  • Add help to argparse etc
  • Full Stops in func comments.
  • Add allData var back into data func
  • Fix importlib model imports for auxillary models
  • Fix output file struture diagram to include logs, checkpoints folders.
  • Add parameter descriptons for LR schedulers in utils.py
  • Echo some of model parameters of config file in gcp_training job
  • Func in notification func that emails status of job if fails, also sends reason for failing.
  • Parse JSON arch utility function
  • Fix gcp hpconfig file
  • Look into training on TPU (https://www.tensorflow.org/guide/tpu)
  • Change staging bucket to bucket in config
  • Remove hard-coded GCP params in config and inject env vars using jq (do this for local psp version as well)
  • Change color of output in training script
  • Look at output suggestions from bandit and make any changes accordingly.
  • Look at output suggestions from flake8 and make any changes accordingly.
  • Add virtual env to workflow (add to readme)
  • Change gcp_notification_func to import secret values from secrets.sh
  • Get job status script
  • Move model layer params to model params
  • A method to create a json config file??
  • Indent optimizer in json to include metaparameters, check if these meta values are set and pass into opimizer function.
  • Input parameter of training script that decides whether to train locally or to GCP.
  • Optimizer tests
  • Re-do config files such that each layer has its individual parameters indented, then pass in via **kwargs...
  • Change main.py to just pass in model-parameters
  • Check to see all config jsons open without error.
  • Upload config file used in model in model folder.
  • Tests_gcp
  • Update build and build status to point to same dir
  • Change filtered to "True" to 1 in configs
  • https://github.com/icemansina/IJCAI2016/blob/master/Train_validation_test_release.ipynb
  • unitest.skip on request URL tests in test_dataset.py
  • Append config fiel to results output file
  • Fix try except in load_dataset.py
  • Output results dont seem t o be working, model logs and metadata not exporting to CSV
  • Remove append_model_output func in utils.
  • Make dummy model simpler
  • Create data dir in psp_gcp
  • in psp_gcp, ensure local training stored in output folder.
  • Change (Keras Model) to Keras.model
  • Change type=5926/6133 to a str, rather than int
  • Add RMSE metric
  • Add save dir to dataset classes
  • If gcp_project!=PRoject then update project
  • Change output_data to output folder
  • Move to new bucket
  • Change structure of network outputs/inputs as https://github.com/wentaozhu/protein-cascade-cnn-lstm/blob/master/cb6133.py
  • Change to TimeDistributed dense??
  • Change all "None" in configs to null
  • Remove name from batch_norm parameter
  • Split up model tests into their own class Test cases .
  • Change "model_" to "model" in dummy json
  • Change Dense_layer1 -> dense_1 in configs
  • Fix order for recurrent layers in Auxillary models.
  • self.assertEqual(model._name, "model_name")
  • Change testLabel -> test_labels in evaluate.py
  • Rename casp10_test_hot to just test_hot
  • Add self to class instance arguments in comments. The self is used to represent the instance of the class.
  • Add TF unit tests
  • Evaluate.py - add raise ValueError if y_true.shape!=y_pred.
  • Add RMSE to plot_history func
  • Try completely removing repeated modules and packages from psp to psp_gcp directories by using the psp dir for the psp_gcp ones as well.
  • Add psp to sys.path so can import from psp_gcp
  • Reset gcp_parameters in config back to ""
  • Update paths for casp10/11 downloads from repo.
  • Add LR scheduler to config parameters.
  • Add input parameter for what callbacks to use.
  • change params = json.load(f) to params = json.load(f)[0]["parameters"]
  • Add output folder name to output_results.csv
  • Change some column names in results file
  • Change where logs are stored in bucket - should be stored in output folder maybe
  • 2 columns of CASP10 precision in results file
  • Add comments to workflow
  • Add reduceLROnPlateau to each config
  • Add references to functions
  • Add repr function to classes
  • Change dataset_size func to size
  • Change model tests to open up each models config and cross-reference with the config values.
  • Change load_dataset to dataset.py
  • Change setup to setUpClass(self) + @classmethod
  • Add API enable commands to psp_gcp gcloud services enable appengine.googleapis.com
  • https://www.tensorflow.org/tutorials/keras/save_and_load#savedmodel_format
  • Add emojis to readme
  • Add banner image to readme