I am trying to train a model in tensorflow.js 1.7, however it seems to require a cuda version 10.0. My host PC is running CUDA 10.2. I want to use a docker container that has tensorflow.js 1.7 installed and working with cuda version 10.0. Is there one that already exists?
i tried creating my own docker image based on nvidia/cuda-10.0
then installing @tensorflow/tensorflow-js-gpu afterwards,
i have the following error, which is complaining of unsupported driver
2020-03-23 02:37:21.103333: I tensorflow/core/platform/cpu_feature_guard.cc:142] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2 FMA
2020-03-23 02:37:21.135247: I tensorflow/core/platform/profile_utils/cpu_utils.cc:94] CPU Frequency: 3991010000 Hz
2020-03-23 02:37:21.135734: I tensorflow/compiler/xla/service/service.cc:168] XLA service 0x50b85a0 initialized for platform Host (this does not guarantee that XLA will be used). Devices:
2020-03-23 02:37:21.135750: I tensorflow/compiler/xla/service/service.cc:176] StreamExecutor device (0): Host, Default Version
2020-03-23 02:37:21.136471: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcuda.so.1
2020-03-23 02:37:21.137060: E tensorflow/stream_executor/cuda/cuda_driver.cc:318] failed call to cuInit: CUDA_ERROR_SYSTEM_DRIVER_MISMATCH: system has unsupported display driver / cuda driver combination
2020-03-23 02:37:21.137085: I tensorflow/stream_executor/cuda/cuda_diagnostics.cc:169] retrieving CUDA diagnostic information for host: 1579792ca4ae
2020-03-23 02:37:21.137094: I tensorflow/stream_executor/cuda/cuda_diagnostics.cc:176] hostname: 1579792ca4ae
2020-03-23 02:37:21.137133: I tensorflow/stream_executor/cuda/cuda_diagnostics.cc:200] libcuda reported version is: 410.129.0
2020-03-23 02:37:21.137161: I tensorflow/stream_executor/cuda/cuda_diagnostics.cc:204] kernel reported version is: 440.64.0
2020-03-23 02:37:21.137172: E tensorflow/stream_executor/cuda/cuda_diagnostics.cc:313] kernel version 440.64.0 does not match DSO version 410.129.0 -- cannot find working devices in this configuration
does anyone have a working docker configuration?
i am on ubuntu 18.04
i have nvidia driver 440 installed with cuda 10.2
i want to use tensorflow.js 1.7 on gpu.
thanks