Sanjoy Das
Sanjoy Das
I don't have enough context here, but it looks like @mjanusz added the override. Michal any ideas? Does the GPU kernel need some changes?
CC @cheshire I agree with both of CK's recommendations.
As the error message says, we don't yet support complex128 on FFT2D. As a workaround can you try using FFT2D with complex64? The right fix is to add support for...
CC @bixia1
JFYI TF 1.4.0 is [very old (released in 2017)](https://github.com/tensorflow/tensorflow/releases/tag/v1.4.0) so I suspect we can't really help you with issues specific to 1.4.0. As for your specific question, can you see...
1.15 is the latest TF 1.x version.
> This is what is not normal, I have enough memory in my system When it says "Unable to allocate 184. MiB", it means that it was not able to...
> I'm getting a ResourceExhaustedError in the middle of a training and I'm assuming that shouldn't be possible. That is surprising, but it is not a logical impossibility. Maybe there...
Hi @Flamefire, Please run the tests as follows: ``` bazel test \ --build_tests_only \ \ --test_env=TF_GPU_COUNT=8 \ --test_env=TF_TESTS_PER_GPU=8 \ --local_test_jobs=64 \ --run_under=//tensorflow/tools/ci_build/gpu_build:parallel_gpu_execute \ \ --nodistinct_host_configuration \ \ --test_tag_filters=gpu,-no_gpu,-nogpu,-benchmark-test,-no_oss,-oss_serial,-v1only,-no_gpu_presubmit,-no_cuda11 \ \...