Model size assumes Fp16 data type

[danielholanda]

Issue

When analyzing a model MLAgility assumes that the model is using FP16 when printing the model size.

On src/mlagility/analysis/status.py lines +115 to +116:

        # Converting number of parameters to MB assuming 2 bytes per parameter
        model_size = model_info.params * 2 / (1024 * 1024)

Possible solutions

• Only show the size in MB once the model has been benchmarked (and a data type has been set)
• Use the datatype of the original model and change the information shown if mlagility modifies the data type (e.g. converts to FP16).

[danielholanda]

@ramkrishna2910 @jeremyfowers