c/c++
use ‘new’ when returning references from functions
keypoint: Dynamically allocated memory is allocated on Heap and non-static and local variables get memory allocated on Stack more from here
tensorflow
solving TF-TRT graph conversion errors
I’ve tried converting a graph in various setups which produced different error messages Original environment ubuntu 16.04 CUDA 9.0.176 cudnn 7.0.x TITAN XP: compute capability 6.1 nvidia driver version 384.130 tensorRT 4.0 original environment + tensorflow-gpu==1.11.0 original environment + tensorflow-gpu==1.12.0 anaconda environment(may not match with original environment) Solution install cudnn Read more…
tensorflow
tensorflow cuda/cudnn support matrix
https://www.tensorflow.org/install/source#tested_build_configurations for some reason, compatability spec is not written for tf1.14. Through experience, I found that it works for CUDA 10.0, cudnn 7.4.2. It could work for higher versions, but I haven’t tested it.
error fix
tensorflow tensorrt import error
the above error shows up whenever I import tensorrt from tensorflow with the following import statement. Environment was: tensorflow-gpu 1.9.0 python 3.6.6 ubuntu 16.04 CUDA 9.0 Solution upgraded to tensorflow 1.10.0. the problem disappeared immediately. This solution was referred in this thread.
deep learning
tensorRT stuff
tensorRT support matrix: https://docs.nvidia.com/deeplearning/dgx/integrate-tf-trt/index.html#matrix to apply the tensorRT optimizations, it needs to call create_inference_graph function. Check here for more details on this function. the graph that is fed to create_inference_graph should be freezed. To know more on what exactly means by “freezing”, check here. for using bare tensorRT python module, Read more…
deep learning
implementing seq2seq RNN model in tensorflow
for tackling the many-to-many(with timedistributed) is easy to implement in keras code, it does not seem to be easy to do so in pure tensorflow. Of course, one can implement it when the sequence length is fixed for all input size. However when the sequence length among batches vary, then Read more…