Training Take Long time

I tried to train the Model On My Desktop Version but till now take more than 13 hours (I keep it running at night while I’m sleeping) , Sometimes the machine is Not Active (I’m going a while and return back) does the Desktop work at the background when the machine is not active e.g.: (screen saver off) ? Shall I change the screen saver to be at least (2 Hours) before turn off?

If your image size is big, you can try to reduce/resize it to smaller size (128x128). That will speed up the training.

Training will continue to run in background as long as your system is awake. You can change your power settings to not to sleep.

Thank You, It works successfully Now with 98% Performance In Validation and training as well, Please check the following link:

I need your help in:

1- How to do the testing with Non-Labeled Data?
2-How to deploy the Model?
3-How I can get confission matrix?
4-How I make the solution after (1 & 2) as Integrated object e.g.: web-service?
5-How to use the model In simulator to do some fun testing e.g.: Enter the Class name , and bring random Image from the same class?

  1. For non-labeled data, you can upload it on the inference/deploy tab. Choose “Dataset Inference” and “Upload” as dataset source.

  2. For deploying the model, goto Deploy tab and choose a run as well as a name and click Reply.

  3. Confusion matrix is not available in Deep Learning Studio yet.

  4. Once you deploy the model, you can do the inference using REST API. A curl example is provided (click on Deployments and launch the deployed app and choose REST API). You can follow the curl example to build a web request in any language you prefer.

  5. I am not sure if I understand what you meant by fun testing. If you meant that you want to try out a random image and let the model predict the class. That is possible by launching the deployed app and uploading the image.

Thank you dear ,

For:

Q.1 , Do you means I provide the csv file with the locations of the Images (without Labels)? please explain cuz no video or documentation provided for this part…
Q.4 , where is the curl example? In addition in the Inference tab there is no Deploy button
Q.5 I mean is there a simulation tool from your experience where I can do some scinario for testing the quality of the algorithm by entering a class name inside text box to get the Images related to this class (similar to google when you search some information and get the Images related to this information)?

In Addition

Q-6 For your video (How to Download And Use Trained Model from Deep Learning Studio) , still not clear how it’s process the video little bit not clear do you have extra documentation?
Also I have Anaconda3 Environment on My Machine , does it do the same work as the Video environment Bash (Ubunto)?

Thank you for your responsive attitude

I think you are using Deep Learning Studio 1.5.x version. Please download the latest version from our website.It supports Deployment/REST API.

Once you have installed the latest version, you can check my answers #2 and #4.

regarding #1, yes. You need to provide a zip file the same way you did for uploading dataset. The only difference is that CSV filename needs to be test.csv and you don’t need to provide labels in test.csv.

regarding #5, there is no simulation tool.

#6, There is no extra documentation. Yes, anaconda3 environment is also fine. Make sure you install the Keras 1.2.2 and other dependencies before running the example inference code.

Can I Install it directly without causing problems to my previous work? I mean shall I run install.exe directly and it prompts that it found earlier version?

From Where I get the latest version?

First you can backup DLS data directory to avoid accidents. The DLS data directory is inside the Deep Learning Studio install folder. You also need to backup <install location>\home\app\database\db.sqlite3

Uninstall the old version and then install the new version in the same location or in a different location. After installation is complete, DO NOT start Deep Learning Studio. First transfer your data folder and db.sqlite3 back to the location where you installed Deep Learning Studio and then start Deep Learning Studio.

You can get the latest version from out website.

1 Like

But It does’t mentions the Version In Your website , My Current Version is 1.5.1 ,

In addition what do you means by (DLS) ?

Yes, it does not mention the version number. I will take that as the feedback. The current version number is 2.0.6 which is on our website.

DLS is short for Deep Learning Studio.

1 Like

But Expect from me some questions after I finished the Installation In case I have some problems

Sorry But All Your Recommendations is not done as you said (I follow it step by step as you said) and I have now to start everything from scratch again, I backed up all the previous setting and return it back to the same setting, I found all the data In the Installed directory but the projects and the database didn’t display when I upload the Deep learning Studio.
Attached copy od the Testing results , the data base location.

I need your help with your support team, It’s difficult to restart all the project again after the whole progress made, please advice.

Please provide the following information:

  1. location of the folder where you backed up old data. Also share the snapshot of the folder

  2. Location of the folder where you install the latest version of Deep Learning studio

Example, if you saved your data folder to c:\X\data and new install location is c:\Y.

To get the old projects back, stop deep learning studio and copy c:\X\data to c:\Y.

After copying, make sure you have c:\Y\data\1\datasets folder and c:\Y\data\1\projects folder present.

I’m not sure whats going On , I think there is a problem , when I return back the previous backup the setting return back to (1.5.1) without the new features in (2.0).

I’m really get upset about this , I did it but the data never goes as it was.

If I put the backup directory (DLS) Then The new Deep learning Studio will load version (1.5.1) without the new features on (2.0.1) that will not work , I think the recommendation of using the Backup (DLS) is not true , I need someone from your team support to contact me to work through how to fix this problem otherwise I have to return back all the previous steps from scratch again , please take this feedback to avoid the same problem with other users of your product.

Thank you , I just restarting the process again , no problem. After finishing the Training I need your help in Video or documentation or guidance in the following points:

1-Testing Non Labelled data (Actual Classification column not Included).
2-Deploying The Model as a web-service.

I used the latest version (2.0.1).

Thank you

I have tried upgrading from 1.5.1 to 2.0.6 on my system after backing up the data folder. I installed on a new location and copied the data folder back. After starting Deep Learning Studio, I found my existing project was missing. So you are right, The steps for backing up and restoring were not complete. There was one more step of copying db.sqlite3 (I have edited my original post with the detail).

However, I didn’t see the second issue you mentioned. After restoring the backup data, I was still seeing 2.0.6 features. Only possibility I can think of is that if you backed up and copied the whole install folder of 1.5.1 version on top of 2.0.6 install folder, then you would see 1.5.1 features.

1 Like

Thank You for your response , And I’m sorry If I looked anger in my writing , but don’t worry I re-process the Model again and it’s on final stages with the same results on (1.5.1) , In any case thank you and your team for this significance product which help others to be more focusing on how to build their models than lost their efforts time on fixing bugs of developing like the case in (R and Python) , your product is even bitter than (Azure ML) and all of us appreciate your dedication and efforts.

Thank you
Mohammad

Ok rajendra can you guide me on how to do the testing for Un-labled data on the current version (2.0.6)?