1. MNIST Trainer

Allows user to upload the MNIST dataset and get a trained model based on the parameters specified. Description about the dataset can be found here.

• URL

  /uploadCompleteScript

• Method:

  POST

• URL Params

  Required:

  width=[integer] : width of images used
  height=[integer] : height of images used
  nClass=[integer] : number of classes
  alpha=[float] : learning rate

• Data Params

  upload=[zip-file] : MNIST dataset linked above.

• Success Response:

  • Code: 200
    Content: {"Accuracy":"[{\"Epoch\": 0.0, \"Accuracy\": \"0.4075\"}, {\"Epoch\": 10.0, \"Accuracy\": \"0.894\"}, {\"Epoch\": 20.0, \"Accuracy\": \"0.8989\"}, {\"Epoch\": 30.0, \"Accuracy\": \"0.9012\"}, {\"Epoch\": 40.0, \"Accuracy\": \"0.904\"}, {\"Epoch\": 50.0, \"Accuracy\": \"0.9105\"}, {\"Epoch\": 60.0, \"Accuracy\": \"0.9012\"}, {\"Epoch\": 70.0, \"Accuracy\": \"0.9113\"}, {\"Epoch\": 80.0, \"Accuracy\": \"0.9156\"}, {\"Epoch\": 90.0, \"Accuracy\": \"0.9146\"}, {\"Epoch\": 100.0, \"Accuracy\": \"0.9143\"}]","trainedModel":"/S3LabUploads/MNIST_data_8.ckpt"}

  Result returned is a json object containing array of epoch accuracy pairs and a relative path to the trained model.

• Sample Call:

  curl -F width=28 -F height=28 -F nClass=10 -F alpha=0.01 -F upload="@MNIST_data.zip" deepc02.acis.ufl.edu:8888/uploadCompleteScript

• Notes:

  You should have "MNIST_data.zip" in the directory from which you are making the above sample CURL call.

2. MNIST Prediction using user models

2.1 : Model Selection

Allows user to upload their own trained MNIST model and later perform predictions using different API(2.2).

• URL

  /generalPredictorModelUpload

• Method:

  POST

• URL Params

  None

• Data Params

  upload=[.ckpt] : A tensorflow checkpoint file.

• Success Response:

  • Code: 200
    Content: Received model : mnist.ckpt

• Sample Call:

  curl -F upload="@mnist.ckpt" deepc02.acis.ufl.edu:8888/generalPredictorModelUpload

• Notes:

  You should have "mnist.ckpt" in the directory from which you are making the above sample CURL call.

2.2 Image Selection

Allows user to perform predictions using the model selected in model section API(2.1).

• URL

  /generalPredictorImageUpload

• Method:

  POST

• URL Params

  None

• Data Params

  upload=[Image-File] : The image file can be of jpeg or png type.

• Success Response:

  • Code: 200
    Content: {"Prediction":"4"}

    Result returned is a json object containing prediction made.

• Sample Call:

  curl -F upload="@img.png" deepc02.acis.ufl.edu:8888/generalPredictorImageUpload

• Notes:

  You should have "img.png" in the directory from which you are making the above sample CURL call.

3. MNIST Prediction on a pretrained model

Allows user to upload MNIST images and get predictions on a pretrained model.

• URL

  /MNISTPredictor

• Method:

  POST

• URL Params

  None

• Data Params

  upload=[Image-File] : The image file should be of 28*28 dimensions. You can find some sample images [here]().

• Success Response:

  • Code: 200
    Content: {"Prediction":"4"}

  Result returned is a json object containing prediction made.

• Sample Call:

  curl -F upload="@img.png" deepc02.acis.ufl.edu:8888/MNISTPredictor

• Notes:

  You should have "img.png" in the directory from which you are making the above sample CURL call.

4. Dashbord Information

Allows user to get information about jobs submitted to the system and their status.

• URL

  /getDashboard

• Method:

  GET

• Success Response:

  • Code: 200
    Content: {"info":{"queriedHost":"127.0.0.1:9042","triedHosts":{},"achievedConsistency":10},"rows":[{"jobstatus":"crashed","jobtype":"training","model":null,"prediction":null,"user_id":"sjs7007testing"},{"jobstatus":"finished","jobtype":"training","model":"/S3LabUploads/MNIST_data_f7cf086e-e63e-44ec-b4af-6caa0764e376.ckpt","prediction":null,"user_id":"sjs7007testing"},{"jobstatus":"finished","jobtype":"training","model":"/S3LabUploads/MNIST_data_8f7b6c4e-c35f-40b8-b89b-c635eadba773.ckpt","prediction":null,"user_id":"sjs7007testing"},{"jobstatus":"finished","jobtype":"training","model":"/S3LabUploads/MNIST_data_cdb8a38d-911c-4d78-8f95-6946f4c0c9fa.ckpt","prediction":null,"user_id":"sjs7007testing"},{"jobstatus":"crashed","jobtype":"training","model":null,"prediction":null,"user_id":"sjs7007testing"},{"jobstatus":"crashed","jobtype":"training","model":null,"prediction":null,"user_id":"sjs7007testing"},{"jobstatus":"finished","jobtype":"training","model":"/S3LabUploads/MNIST_data_8ea199e2-955d-42ee-b60e-4a56b11f297f.ckpt","prediction":null,"user_id":"sjs7007testing"}],"rowLength":7,"columns":[{"name":"jobstatus","type":{"code":13,"type":null}},{"name":"jobtype","type":{"code":13,"type":null}},{"name":"model","type":{"code":13,"type":null}},{"name":"prediction","type":{"code":13,"type":null}},{"name":"user_id","type":{"code":13,"type":null}}],"pageState":null}

  Result returned contains an array of json consisting of following information :

  Job status : finished/crashed
  Job type : training/testing
  Prediction : in case of a testing job, else null
  User_id : user who submitted the job
  job_id : unique job id pid : process id assigned to job

• Sample Call:

  curl deepc02.acis.ufl.edu:8888/getDashboard

5. Dashbord Information : Specific

Allows user to get information about jobs submitted by specific user to the system and their status.

• URL

  /getDashboardSelective

• Method:

  GET

• URL Params

  user_id : user_id whose jobs you want to manage.

• Success Response:

  • Code: 200
    Content: {"info":{"queriedHost":"127.0.0.1:9042","triedHosts":{},"achievedConsistency":10},"rows":[],"rowLength":0,"columns":[{"name":"pid","type":{"code":13,"type":null}},{"name":"job_id","type":{"code":12,"type":null}},{"name":"jobstatus","type":{"code":13,"type":null}},{"name":"jobtype","type":{"code":13,"type":null}},{"name":"model","type":{"code":13,"type":null}},{"name":"prediction","type":{"code":13,"type":null}},{"name":"user_id","type":{"code":13,"type":null}}],"pageState":null}

  Result returned contains an array of json consisting of following information :

  Job status : finished/crashed
  Job type : training/testing
  Prediction : in case of a testing job, else null
  User_id : user who submitted the job
  job_id : unique job id pid : process id assigned to job

• Sample Call:

  curl deepc02.acis.ufl.edu:8888/getDashboardSelective?user_id=sjs7007

6. Kill Jobs

Allows user to kill training jobs with specific job_id.

• URL

  /killProcess

• Method:

  POST

• URL Params

  Required:

  job_id=[UUID] : job_id of job.
  pid=[integer] : process id of the child process spawned to perform training/testing.

• Success Response:

  • Code: 200
    Content: Killing process with PID : 12291

• Failure Response:

  • Code: 400
    Content: Process killing failed : not a child process

• Sample Call:

  curl -F job_id=c8c2e97c-c8b1-4c6c-ae0f-8d7c6d0ceee6 -F pid=12291 deepc02.acis.ufl.edu:8888/killProcess

• Notes:

  You can only kill processes which are generated by this application.

7. Suspend Jobs

Allows user to suspend training jobs with specific job_id.

• URL

  /suspendProcess

• Method:

  POST

• URL Params

  Required:

  job_id=[UUID] : job_id of job.
  pid=[integer] : process id of the child process spawned to perform training/testing.

• Success Response:

  • Code: 200
    Content: Suspended process with PID : 12291

• Failure Response:

  • Code: 400
    Content: Process suspension failed : not a child process

• Sample Call:

  curl -F job_id=c8c2e97c-c8b1-4c6c-ae0f-8d7c6d0ceee6 -F pid=12291 deepc02.acis.ufl.edu:8888/suspendProcess

• Notes:

  You can only suspend processes which are generated by this application.

8. Resume Jobs

Allows user to resume training jobs with specific job_id and pid.

• URL

  /resumeProcess

• Method:

  POST

• URL Params

  Required:

  job_id=[UUID] : job_id of job.
  pid=[integer] : process id of the child process spawned to perform training/testing.

• Success Response:

  • Code: 200
    Content: Killing process with PID : 12291

• Failure Response:

  • Code: 400
    Content: Process killing failed : not a child process

• Sample Call:

  curl -F job_id=c8c2e97c-c8b1-4c6c-ae0f-8d7c6d0ceee6 -F pid=12291 deepc02.acis.ufl.edu:8888/resumeProcess

• Notes:

  You can only resume processes which are generated by this application.

9. Test on trained model

Allows user select an already trained model based on job_id and perform testing using that model.

• URL

  /testTrainedOnline

• Method:

  POST

• URL Params

  Required:

  job_id=[string] : job_id of model to be used.

• Data Params

  None.

• Success Response:

  • Code: 200
    Content: {"Prediction":"3\n"}

  Result returned is a json object prediction value.

• Sample Call:

  curl -F job_id=dc725ed7-dc51-4c9a-9057-f1539a958681 localhost:8888/testTrainedOnline -F upload=@generalPredictor/93.png