Closed gaborvecsei closed 3 years ago
@gaborvecsei @reedwm I am sorry for late reply. Just noticed this issue...
I think the issue is because https://github.com/tensorflow/models has been refactored such that the official.benchmark
module only exists in the benchmark
branch but not the master
branch.
This issue can be fixed by running PerfZero with --git_repos="https://github.com/tensorflow/models.git;benchmark"
so that PerfZero will checkout the right branch of this repository.
@gaborvecsei, could you re-try after making the change described above? I will update the README afterwards.
@lindong28 Thank you for your response! Now I can run the benchmark. I only tried with the following setup:
python3 /workspace/perfzero/lib/benchmark.py \
--git_repos="https://github.com/tensorflow/models.git;benchmark"
--python_path=models \
--gcloud_key_file_url="" \
--benchmark_methods=official.benchmark.keras_cifar_benchmark.Resnet56KerasBenchmarkSynth.benchmark_1_gpu_no_dist_strat
@gaborvecsei Thanks for the confirmation! The PerfZero README has been updated to use the right repository link.
I wanted to run some benchmarks on my local machine and the problem happened to me which you can see at the title.
python3 benchmarks/perfzero/lib/setup.py --tensorflow_pip_spec=tensorflow==2.3.0
docker run -it --rm --name tf-test -v $(pwd):/workspace --gpus all perfzero/tensorflow bash
I can see that https://github.com/tensorflow/models does not have
official.benchmark
module, but then how should I run the benchmark? Could you update theREADME.md