Hi Derek!
Do you mean completely dropping it or keep it in those Quantizer/Serializer/Converter tools and only dropping it for runtime? One of my more recent workflows was we needed a quantized PoseNet, the goal was not really accuracy but mainly speed. It was a float model trained in TF, so one of the ideas was to convert it to .armnn directly from TF and perform quantization using Quantizer and run the .armnn quantized model afterwards.
I would agree with dropping support for runtime if it would free significant resources, because I don't see a reason why to run caffe or TF on embedded, but I would not be so sure to drop support completely for an offline approach. Otherwise there still might be some issues with op-support in conversion between TF/TF Lite, so using TF instead of TF Lite is still viable. The usage of offline tools on the other hand shifts the design more towards things such as offline ML compilers or other offline tools and funnel everything through an optimized output, which in Arm NN's case would be .armnn.
Best Regards, Pavel
-----Original Message----- From: Armnn-dev armnn-dev-bounces@lists.linaro.org On Behalf Of Derek Lamberti Sent: Friday, February 21, 2020 9:45 AM To: armnn-dev@lists.linaro.org Subject: [EXT] [Armnn-dev] Tensorflow and Caffe parser deprecation
Caution: EXT Email
Hi all,
I would like to test the communities appetite for deprecating the Tensorflow and Caffe parsers. This would free-up some dev and test resource to focus on potentially more relevant features. The .armnn, .tflite and .onnx formats would continue to be supported and actively developed as viable alternative routes into ArmNN.
I would be interested to know if you think this move would significantly, negatively affect any known/existing workflows with ArmNN. Any thoughts or comments are welcome.
Thanks, Derek IMPORTANT NOTICE: The contents of this email and any attachments are confidential and may also be privileged. If you are not the intended recipient, please notify the sender immediately and do not disclose the contents to any other person, use it for any purpose, or store or copy the information in any medium. Thank you. _______________________________________________ Armnn-dev mailing list Armnn-dev@lists.linaro.org https://eur01.safelinks.protection.outlook.com/?url=https%3A%2F%2Flists.lina...