Vladimir Kuzmin
Vladimir Kuzmin
### Summary: Improve current fat jar to make it completely self-contained JAR with OS-Arch specific native libraries. New "fat" JAR has embedded resources with JNI library, Neuropod Core and Backend...
## Bug - Using OPE: no - Neuropod backend TensorFlow, CPU Found during benchmark that in IPE mode TF backend reports warning on every request that itself affects performance: neuropod/backends/tensorflow/tf_tensor.cc:140:...
## Feature Add metadata attribute "native platform version" as indirect specification for model's data format and platform that was used to generate data. Tensorflow and Torchscript models has binary data...
## Bug New env. var NEUROPOD_BASE_DIR allows to specify location where register backend searches for backends. I tried to update NEUROPOD_BASE_DIR from parent process at the very beginning (used setenv...
It seems that PyTorch models in Neuropod universe allows some "variant reading". This is a result of historical reasons when neuropod didn't have separate Backend projects and user could care...
## Feature Support fat, fully self-container Neuropod API jar with OS-Arch specific resources that has JNI lib, Neuropod Core and TF, Torchscript backends. Note that currently Bazel build has target...
I looked at "Best practices for using the Java Native Interface" https://developer.ibm.com/technologies/java/articles/j-jni/ and found reasonable advises: 1. "For new JNI code validate that there is a check for an exception...
## Feature C API We don't expect that somebody is using C API directly. This will be used by future CGo API mostly. Connected to https://github.com/uber/neuropod/issues/294 ## Describe the solution...
I suggest to add Support section to README.md that has a link to Issues and stackoverflow ``` ## Support Use [stackoverflow](http://stackoverflow.com/questions/tagged/neuropod) and/or [Github issues](https://github.com/uber/neuropod/issues). ``` If agree, feel free to...
## Feature Currently neuropod.backends.keras.packager and neuropod.backends.tensorflow.packager supports Tensorflow 1.x interface. It should provide support for Tensorflow 2.x, ideally it should be generic interface that allows client to work with 1.x...