I am looking at using pytorch and TorchScript as the DL framework for my company. One of the datapoint we were looking for was to check if there are companies (other than FB) which have used TorchScript/C++ already in production. Could not find info with searching on the internet. Any inputs would be appreciated.
@Michael_Suo Would you like to chime in on this one?
If you want to use C++ then you’ve got no choice, TF C++ API is dead. If you mean pure inference TF is better choice.
Selecting framework would mean fx should be good at (1) Model development (2) Production serving.
For model development, I think python is a reasonable choice (inspite of GIL issue it has).
For production inference TorchScript is a solution to productionize the model, so wondered if some companies/people already tried it out and what the findings were.
Hi @Param, I am curious as to how you finally proceeded and what was the deciding factor? I am in a similar situation right now and could use some advice.