Struct surrealml_core::execution::compute::ModelComputation

pub struct ModelComputation<'a> {
    pub surml_file: &'a mut SurMlFile,
}

A wrapper for the loaded machine learning model so we can perform computations on the loaded model.

Attributes

• surml_file - The loaded machine learning model using interior mutability to allow mutable access to the model

Fields

surml_file: &'a mut SurMlFile

Implementations

impl<'a> ModelComputation<'a>

pub fn input_tensor_from_key_bindings( &self, input_values: HashMap<String, f32> ) -> Result<ArrayD<f32>, SurrealError>

Creates a Tensor that can be used as input to the loaded model from a hashmap of keys and values.

Arguments

• input_values - A hashmap of keys and values that will be used to create the input tensor.

Returns

A Tensor that can be used as input to the loaded model.

pub fn input_vector_from_key_bindings( &self, input_values: HashMap<String, f32> ) -> Result<Vec<f32>, SurrealError>

Creates a Vector that can be used manipulated with other operations such as normalisation from a hashmap of keys and values.

Arguments

• input_values - A hashmap of keys and values that will be used to create the input vector.

Returns

A Vector that can be used manipulated with other operations such as normalisation.

pub fn raw_compute( &self, tensor: ArrayD<f32>, _dims: Option<(i32, i32)> ) -> Result<Vec<f32>, SurrealError>

Performs a raw computation on the loaded model.

Arguments

• tensor - The input tensor to the loaded model.

Returns

The computed output tensor from the loaded model.

pub fn buffered_compute( &self, input_values: &mut HashMap<String, f32> ) -> Result<Vec<f32>, SurrealError>

Checks the header applying normalisers if present and then performs a raw computation on the loaded model. Will also apply inverse normalisers if present on the outputs.

Notes

This function is fairly coupled and will consider breaking out the functions later on if needed.

Arguments

• input_values - A hashmap of keys and values that will be used to create the input tensor.

Returns

The computed output tensor from the loaded model.