As a "feature request", I was wondering if we could get some runtime statistics about running decoding with either ipex or openvino or others runtimes ?
Is that available somewhere already (I couldn't find it) ? I'm trying to decide if it's worth investing into making ipex work in our code, but I was wondering what kind of gains to expect with the conformer with ipex with or without quantization. Openvino looks interesting too, but that again is a non negligible amount of work.
As a "feature request", I was wondering if we could get some runtime statistics about running decoding with either ipex or openvino or others runtimes ?
Is that available somewhere already (I couldn't find it) ? I'm trying to decide if it's worth investing into making ipex work in our code, but I was wondering what kind of gains to expect with the conformer with ipex with or without quantization. Openvino looks interesting too, but that again is a non negligible amount of work.