Collaborate on Invoice Template AI for Facilities with Ease Using airSlate SignNow

Invoice template ai for Facilities

hi friends welcome to my channel so today we are going to learn about UAP paath AI Center uh how we can navigate to this AA cender how we can build ml skill in this AA cender which we can use in our RPA workflow uath workflow uh to use in extractors classification how we can use that ml model so today we will be learning about how we can make this ml skills ml model from the AA Center so uh we can access the AA center from our uath Cloud platform so here when we so I'm using the new navigation so there are slights uh slights of changes um in the UA so here we will be able to see as and actions and uh documentation document understanding orchestrate uh everything but sometimes uh we will not be able to see so for that we need to uh enable the particular product from the services so for that we need to go to the admin and we need to take that uh tenant and when we go to the services we will be able to find um the products which are there so we need to click So currently I have enabled so we can from here we can enable this and when we launch it will open and after that we will be able to see the all the products which we have enabl from the services in this uh left side portal so let's go to the AA Center so today we will be uh going to break down of the main steps and fe uh features that we would typically use in AI Center for projects so first we if we have already created projects we will be able to find um that projects in uh this my projects field So currently I have created one test so um we are able to see it here and to create project we need to click create and we need to enter a name for our project um so something demo something like that and you can give uh the project description it's an optional and you can click like restrict assess uh something as required and need to click create and our demo project is created so here we will be able to uh find um the steps which we were going to use so dashboard so in dashboard we will be able to find the overall details of this uh project and data set so data set means um for training and evaluating ml models we request U data sets right so from here we will we can create the data set or we can even upload the folders which contains the data sets data uh data sets so today we are going to learn about uh how we can um train an existing ml model in the AA cender which is retrain uh retrain model for invoices and how we can create the ml skill from that model so the data set is for that uh just to create the data set for uh our project and uh next is data labeling so data labeling is simply to label like uh if we are using the ml skills for uh extracting data from receipts invoice such like that cases we need to label the data which uh the board need to fetch right from uh so for that we are using this data labeling just to label the data so here you will be able to see uapa document understanding and choose from oob label templates so this will create a session with our uapa document understanding label interface and here it will uh create a session with labeling interface for data labeling from out of the box labeling templates so here we use document understanding labeling interface and here it's using out of the box labeling templates and ml package so here we will be able to upload a new package to the AI cender or we will be able to see uh the uh ml packages which are already there so if we need to add a new ml package we will upload here uh by use converting that ml uh package into a zip file we will upload it here like we need to uh enter the name of that package we need to upload the package the zip file um and it's input type uh like how it's files or Json like that and if any we can provide this is optional about the input description and what output we are expecting and the language so uh it should be in Python only python is supporting and like we can enable the custom package version and we can give the package version which required and if it's for training we can enable training uh and can enable uh recommend GPU if we uh GPU is enabled it will work fastly actually and if I want to use an ml package developed by the UA path itself like uh document understanding or any open- Source Community we use this out of the box packages okay and uh we use this uh when we need to import an ml package that was exported from AA CER previously so we have uh previously uh after training and all we have exported an ml package and we need to use that we use this So currently as uh I said we are using an ml package which is developed by the UAP paath for document understanding so let's take out of the box packages so here you will be able to find a lots of packages which is already created by uath which is retrain and not retrain like path understand document understanding image analysis language compr uh comprehensions language analysis for image an analysis so different so uh we will learn uh about the both like uath image analysis and also uath document understanding so first let's uh make an ml model for invoices as I said so it it's like need to extracting the details using the ml from the documents right it's in it will be coming in document understanding so we need to use uath document understanding and here also we are able to see lots of M uh ml packages which we can use uh for our for like our nit so here there are invoices ID cards if we need to fet from ID card we can use this invoice India invoices okay so uh I will take this invoices so here when we go it's retrain so we can retrain this uh ml model uh like how we needed right so it's package is 20 we are taking the latest package only and let's click submit so we need to give an uh name for our ml pack package so I will give some um invoice and the description input output are the if you need to change we can change that and let's click create so now it status is undeployed so now we need to uh deploy so for that first we need to label rate so let's go to the data labeling and uh we need to create a new labeling session for training uh it means to get what all we are needed right to make uh to retrain that ml model that invoice model so we need to give a name invoice and if a data table is already there we can choose from here means after creating uh sorry data set so we can create here or we can upload so we can use that here uh or we can create from here also so uh need to give a name and we can give description can make data set public or private like that so the data labeling is deploying sometimes it will take some time to deploy so it's now it status is changed to available so this data labeling section is available so when we click that we will go insert this data labeling so now we need to what do what we need to delay able uh the invoices like what for what it will be coming so for that first we need to import the files which we are using to train the model so need to click the import and here we can give the name for the batch I'm giving like batch one and click to upload or drag drops so I have made some invoices I will show you that I'm importing that click upload so 20 files uploaded s successfully then we need to click import so it's importing the status so if it fail if in in any step it's fail and if you need to know why it's failed every time it will be in this um ml logs for every success whatever we done the steps will be loog logging in the uh ml logs and we will be able to find it there so let it import and also here also we can see the logs about the importing status so all the files imported successfully so let's close this so we will be able to find that files here so now we need to do what we need to uh we need to train the model like which all fields we needed so for this uh to create the fields we use regular uh fields and classification Fields so let me show how it's work for regular field we can give some some values I will give like invoice number okay uh content type is string shortcut is there okay save so this is the invoice number and uh we can give date it's date right so I will give like date and what Bild to right we can give Bild to build TST string and can give toal let's give it as number and what can what else can give can give account number we can take details from the table also easily account number account number is number right then can give like something else transferred account to which if we use uh the already invoice invoice uh invoice uh package which is the UA before returning we will not get exactly like what um Regular field names we need to give like that it will fetch like because it's in its trainable form right so we need to retrain to like how we Reed number okay now let is classification field classification field is like if any currency or something if we need that here if we need column values we will click here actually I I don't see that so we need to get the column Fields here we require this classification field is actually for classifying so it's uses like Indian maybe sometimes it will come like Indian rupee or USD something like that so if it's change we need to we use like that so now we will not do this as it's not required so here if we need to get the column any column we will create it here so let's create description string right save then days days is number then we'll give unit price number or uh string anything can give if you need uh rupes also with that so we have created the regular fields and the column fields which we need to fetch so let's retrain that so invoice number is this right so let's click this and click here so we have uh Mark um means so we are labeling it it so we have labeled the invoice number so let's label the date so we have labeled the date let's label Bill to let's build total let's build account number and let's build transfer account like GA only wrate account number okay then let's now now do this description so let's take both and can give like that so we will be able to find it that two columns here in days so just we can take like this unit price then amount okay so we have done for one so we need to do for the the like this we need to do for the 20 things but instead of we do this manually when we click this predict the as Ander use generate A8 to predict automatically to the corresponding names now you are seeing right but for this it should be always uh always we should give correct names so it's fetching correctly right by using predict in so instead of manually we do when we click the predict the a sender use generate a to predict automatically so when you go here also and click predict so here there are also like generative pred with the generative model so here it's combine generative with pre-labeling model so you can choose like that and click predict so instead of we manually label each when we click predict uh the uh generative AA will automatically predict the corresponding value and if it's not fine we can change so like that we can do for a every docks so after we label all the documents now we need to export this data to the data sets by clicking this export so let me click export and here we need to give an name uh for this export so let us let me give like invoice or something invoice and here we can click like all labeled that labeled only schema like what we need to export so I have clicked all and let's uh click export to AA or we can download this exported schema to the Excel or we can download it as file but let's export to the AA Cent so it's exporting the each files so we can see that here so let it finish the export thing is completed now let's go back to our data labeling AA cender and when we go to the data set this is the data set which we created right so here we will be able to find the exported thing which we created now so we are able to export the data we label to this uh data set here so uh we have completed the data labeling and we have exported the data to the data set now what we need to do we need to uh create an ml skill which we can use in our uh process to use the uh ml package which we have retrained and this uh here it's uh pipeline is there this pipeline we used to um run the uh we have train right retrain right uh right uh the ml package to run that we use this uh pipeline so let's create ml skill so need to give one name I will give like ml skill invoices then need to choose the package this is the package we created right and the major version and choose uh package minor version we need if we need to give any description we can give and this will make the skill public or something and to auto update we can enable this and if you need to undeploy the skill uh within some days we can choose this and if Advanced infra stressing means uh if we enable GPU enable GPU is will what will train this uh ml skill fastly for that only so let's [Applause] create and this replica amount can give two now let's click create so it's it it's deploying only it will take time so now we need to run the ml model we created right so for that we need to create the pipeline it will take some time to change the status so before that uh we need so actually this will deploy with the old data which is not retrained so for retraining we need to create the pipeline to train the model with the new data label we have done so for that need to click the create new Pipeline and here we can take full pipeline or if we need only to train we can use train run or we need to evaluate we can use evaluate TR so I will choose train run only and can give name here invoice p typ line and choose a package we will be able to find the package and choose major version this and the minor version this and the input data which we created so we need to use that you need to click this export which we created and enable GPU will do it fastly and let's SC it so here if we click run now it will run or we can schedule the pipeline run also when we need it so here I have choose run now so the pipeline is running the bo uh the package invoice package is getting trained with the new data set we have uploaded we have created by data labeling so it will take some time so after that only we need to create the ml skill so actually I have this will this will be deploying uh with the uh ml skill which already there invoice because we haven't trained right after completing this running and and if when the pipeline status become success it means uh the training is completed and after that we we can create the ml skill with that ml package which we have trained and we we can use that ml skill in our processes to extract uh the invoice details details uh to classify the invoice like that so it's running only it will take time so let it complete so here you can see this ml skill is now available so you will be able to find this ml skill in your process um in your process when you extracting means when you connected um your process uh with the same tenant you will be able to find the ml skill and you can use there or if you have created it at public you will be able to find the uh URL also from here here right here [Music] somewhere actually I think I haven't created it at public so that if it's public we will be able to find the url also so you can place that URL or you can if it's connected to the same tenant you will be able to find this ml skill in the extractor and you can use that to extract the details so still now pipeline is running that means the boat um it's uh the ml package is getting trained with the new data set which we have provided so let it complete so after it complete it status will be available and like how we have created the ml skill we can again create with that trained ml package and you will be able to find the score of this training uh the pipeline run so with the score maybe you can check whether again need to retrain or how what to do next because in sometimes the boat will not be able to extract correctly from the invoice with the current uh ml model so we can use retrain uh activity even for classification and extraction to retain the model by checking the score checking the value of this threshold like how we need so you can see here it as successful so uh actually at first it was failing because we have we haven't taken that data set correctly so I will show again but it will take a lots of time to train um so we need to give the invoice name and we need to choose uh the package and the major version and now you will be able to see uh the minor version as one also because already we have re uh retrained right and it's successful so it's uh version updated to one so you can choose that or else if you need to uh train the same model then choose zero and while choosing data set we need to uh take uh this only not the export or uh before that we need to take the uh data set we need to uh uh take uh this only not the export or uh before that we need to take the thing inside the export so first time I haven't choosed that because of that the training was getting failed then we need to create it's failed okay to the pipeline created successfully so it will run and after that it will become successful and you will be able to find the score So currently the score is one so it means that it's uh 100 percentage correct create the ml skill again with the uh trained uh new package so when we go to the ml skill and need to create new and can give name uh then need to choose the package and when we um go to the minor version we will be able to find zero which is the version which which was before retraining and one after retraining and here we can give like undeployed skill after period like uh have said earlier um we can select that like we needed and can enable or to update make ml skill public like that then need to click create at least two replica so it was showing one warning like replica con should be minimum too so I have increased that sorry uh so already there is an ml skill with the same so we can see here the version likes 24.44% ml skill uh in the uh uath processes we will be able to use this to extract the data from the invoices so here uh the score level is less uh if the score level is s we can again retrain the ml skill and based on the uh score we can uh check whether how much score is it and again retrain or can do like that