Closed baroai closed 11 months ago
When I checked the commit version 559f08f
I was able to identify the error. The error was related to the inability to import 'variable_v1,' and it indicated that the model's library could not be found in TensorFlow versions 2.5 and above. I made the same changes to this part as I did in my previous work, for example, changing variable_v1
to variables
and I confirmed that GPU operations work afterward. I would like to know why the variable name was changed to one that doesn't work like this.
The master branch of this repo has been updated to work with the latest nightly version of TensorFlow, and so it no longer works with older versions like TF 2.7.0. You can try an older commit, such as one before 559f08f2cc76c9dbd30f25abc66acc516d1b4bd0.
I git this error with
GPU = 4090 RTX 2ea CPU = Intel OS = Ubuntu20.04 Tensorflow=2.7.0 Protobuf=3.20.0 python=3.9.0
run command is
python tf_cnn_benchmakrs.py --model resnet50 --batch_size 200 --num_gpus 2
and i got this errorprevieous tf_cnn_benchmark is running correctly but now it doesn't work...
what should i do?