Multiple model inputs for BlindAi

[kbamponsem]

Description

Implementation of multiple model inputs. This request contains updates on both the server and client.

• We add a new message to the protobuf file

  message TensorInfo {
  repeated int32 fact = 1;
  DatumTypeEnum datum_type = 2;
  }

  This represents an object capturing both the input_fact and datum_type of an input.

• Also, a new method is added to model.rs file

  fn create_inputs_for_tensor<
  A: serde::ser::Serialize + tract_core::prelude::Datum + serde::de::DeserializeOwned,
  >(
  input: &mut [u8],
  ) -> anyhow::Result<Vec<Vec<u8>>> {
  let inputs_for_tensor: Result<Vec<Vec<u8>>>;
  let input_vec: Vec<Vec<A>> = serde_cbor::from_slice::<Vec<Vec<A>>>(input)?;
  inputs_for_tensor = input_vec.into_iter().map(cbor_get_vec).collect();
  inputs_for_tensor
  }

  This method deserializes and serializes the Vec<Vec<A>> passed from the client.

• For load_model, since we can have multiple inputs, we accumulate the different inputs, their indexes and datum_type information before storing the model in InferenceModel

      for (idx, tensor_input) in tensor_inputs.clone().iter().enumerate() {
          let mut input_fact: Vec<usize> = vec![];
  
          for x in &tensor_input.fact {
              input_fact.push(*x as usize);
          }
          datum_input = convert_type(tensor_input.datum_type.clone())?;
          datum_outputs = tensor_outputs
              .iter()
              .map(|v| convert_type(*v).unwrap())
              .collect();
          datum_inputs.push(datum_input.clone());
          input_facts.push(input_fact.clone());
          model_rec = model_rec.with_input_fact(
              idx,
              InferenceFact::dt_shape(datum_input.get_datum_type(), &input_fact),
          )?;
      }

• For the client, the interface changes to accommodate the multiple inputs

  
  tensor_inputs=[[(1, 480, 480, 3), ModelDatumType.F32]]
  tensor_outputs=[ModelDatumType.F32]
  client.upload_model(model, tensor_inputs=tensor_inputs, tensor_outputs=tensor_outputs, sign=True)

Now data_inputs can either but a List or List[List], depending on whether the model is a multiple input model.

client.run_model(data_inputs, sign=True)

## Related Issue

This project accepts pull requests related to open issues. If suggesting a new feature or change, please discuss it in an issue first.
If fixing a bug, there should be an issue describing it with steps to reproduce.
Please link to the issue here.

## Type of change
The types of changes must be deduced from the labels of the related issues. Please consider providing these further details.

- [x] This change requires a documentation update
- [x] This change affects the client
- [x] This change affects the server
- [x] This change affects the API
- [ ] This change only concerns the documentation

## How Has This Been Tested?

It has only been tested with integration, but not with unit tests. Need to update the unit tests directory with the latest from the `master` branch.

## Checklist:

- [x] My code follows the style guidelines of this project
- [x] I have performed a self-review of my code
- [x] I have commented my code, particularly in hard-to-understand areas
- [ ] My changes generate no new warnings
- [ ] I have updated the documentation according to my changes

<!--- To add when we set tests-->
<!-- 
- [ ] I have added tests that prove my fix is effective or that my feature works
- [ ] New and existing unit tests pass locally with my changes
-- >

[JoFrost]

Thanks for your recent changes! I will do some tests on my side. Let's get this merged asap :)

[kbamponsem]

Thanks for your recent changes! I will do some tests on my side. Let's get this merged asap :)

Alright. Noted!

[github-actions[bot]]

☂️ Python Cov

current status: ✅

Overall Coverage

Lines	Covered	Coverage	Threshold	Status
392	347	89%	0%	🟢

New Files

No new covered files...

Modified Files

File	Coverage	Status
client/blindai/client.py	91%	🟢
TOTAL	91%	🟢

updated for commit: 225aab6 by action🐍