partial fc petrained model inference: what to use for fp16 flag

I am trying to make an inference using the partialfc r100 and r50 models, and I have downloaded them from:
https://github.com/deepinsight/insightface/tree/master/recognition/partial_fc#5-pretrain-models

I see that the models are named `16_**` and `16*`

Is it safe to assume that these are fp16 models?

So, in line:

https://github.com/deepinsight/insightface/blob/17cdeab12a35efcebc2660453a8cbeae96e20950/recognition/arcface_torch/inference.py#L22

should I be setting it as:

```
net = get_model(name, fp16=True)
```

is this the correct way to set the model for inference on partialfc using the given pretrained model?

thank you for the awesome package.