Hi, I have deployed TVM in my projects and the result is amazing! Especially when we applied auto-tuning. But I have a question: once the network is auto-tuned, when we add more data to finetuning this model, should we auto-tune it again? The model symbol is the same as the previous one, only params are changed.
Not necessary unless your model structure or shape changes. Auto-Tuning is for operator not the entire model.
Got it! Thanks~:grin: