anrgusc / defer Goto Github PK

Hi,
I am trying to reproduce the results following your guidance.

First, I run 'local_infer.py', everything works well and the code downloads an "resnet*.h5".

Then, I set computeNodes as [ip1, ip2]. I also modified part_at as ["conv3_block1_add"].

I run 'node.py' in ip2 node. And finnaly I run "test.py" in ip1 node.

But I met some problem:

Exception in thread Thread-1: Traceback (most recent call last): File "/home/anaconda3.6/lib/python3.6/threading.py", line 916, in _bootstrap_inner self.run() File "/home/anaconda3.6/lib/python3.6/threading.py", line 864, in run self._target(*self._args, **self._kwargs) File "../src/dispatcher.py", line 118, in run_defer self._dispatchModels(models_to_dispatch, self.computeNodes) File "../src/dispatcher.py", line 61, in _dispatchModels print("getting weights: ", models[i].get_weights()) File "/home/anaconda3.6/lib/python3.6/site-packages/tensorflow_core/python/keras/engine/training.py", line 173, in get_weights return super(Model, self).get_weights() File "/home/anaconda3.6/lib/python3.6/site-packages/tensorflow_core/python/keras/engine/base_layer.py", line 1352, in get_weights return backend.batch_get_value(params) File "/home/anaconda3.6/lib/python3.6/site-packages/tensorflow_core/python/keras/backend.py", line 3185, in batch_get_value return get_session(tensors).run(tensors) File "/home/anaconda3.6/lib/python3.6/site-packages/tensorflow_core/python/client/session.py", line 956, in run run_metadata_ptr) File "/home/anaconda3.6/lib/python3.6/site-packages/tensorflow_core/python/client/session.py", line 1105, in _run raise RuntimeError('The Session graph is empty. Add operations to the ' RuntimeError: The Session graph is empty. Add operations to the graph before calling run().

It seems like the tensorflow version might cause the issue. I am using tf-1.15.0.

May I know which tf-version you are using? Or did I make some mistakes during the experiments?

Many thanks.

Hi,

When I tried to follow, I was confused about the content of “For each model, we varied the number of compute nodes between 4, 6, and 8. This was done using the CORE Network Emulator (http://coreemu.github.io/core), which allowed us to create network topologies simulating these
different node configurations and run the simulation locally in a close-to-zero latency environment.” in your paper. Do I need to download the CORE Network Simulator on a Linux system and simulate computing nodes on it for experimentation? I tried using the server in the laboratory as the main node and my desktop as the computing node. However, even if the server and desktop can establish communication and the IP parameters in the code are correct, the program still cannot run.

I would greatly appreciate it if you could provide an answer.

Hi, I was recently reading this paper of yours and I saw that you used CORE to simulate the number of nodes to be calculated, but in this paper of yours you used python 3.6, doesn't CORE require the use of python 3.9+? What should I do with the different requirements of the two python versions?