Closed kit-high closed 1 year ago
Hi @kit-high, thanks for your interest in CMSIS-NN. In general we add optimizations for operators that are supported by TFLM as we are an optimization library and not a ML framework. You can expect new operators like transpose conv and new data types like int4. Is there any particular operator you are interesting in?
@mansnils I apologize for the delayed response.
I agree with you that it's not an ML framework. However, since it already has implementations for somewhat complex ML methods like LSTM, I was expecting that it would also implement methods that are not yet included, such as RNNs or Transformers.
As an engineer, I feel that the advancement of Light Edge AI is closely tied to the development of this library. If the opportunity arises, I would like to contribute by making a pull request. Thank you.
Thanks @kit-high. Your contribution is welcome. Anyhow a potential new operator should have an equivalent reference kernel in TFL/TFLM.
Hi, thank you for your work!
Is there a future operation roadmap in long/short term like adding RNN... transformer...?