can't run baseline_rl.py

[cihantugrulcicek]

I wasn't able to run the baseline_rl.py. When I run the code with from_file=True, and providing x_dir and adj_dir,

• First, I got a file not found error for the adj file. I guess this occurs as adj files start with adj- in the validation set. However, this is not specified in batch_env_rl.
• Even though I rename the adj file with the same name of the x file, I got the following error on line 44 of 'batch_env_rl',
  'ValueError: could not broadcast input array from shape (20,3) into shape (50,3)', which seems that the number of nodes are not read from file, instead, read n_nodes argument in the constructor.
• After passing 'n_nodes' argument with 'from_file=True', I still got the error Sizes of tensors must match except in dimension 1. Got 32 and 1 in dimension 0 (The offending index is 1) on line 54 in 'train_baseline.py'.

[paulorocosta]

Hi @cihantugrulcicek,
Thanks for bringing this to our attention.
I have updated the code to read from_file=True when using the batch_env_rl.
An example of how you can call it can be found below. Note that when reading files with a different number of nodes, the instance features are padded with zeros when the number of nodes is smaller than the maximum number of nodes in the files.

    import os
    base_dir = os.path.dirname(os.getcwd())
    x_dir = os.path.join(base_dir, 'data', 'valid', 'instances')
    adj_dir = os.path.join(base_dir, 'data', 'valid', 'adjs')
    batch_env = BatchEnvRL(n_envs=2, from_file=True, x_dir=x_dir, adj_dir=adj_dir)

Please let me know if this is working now.

Thanks again for your feedback!

[cihantugrulcicek]

it works now.. thank you for quick respond

[cihantugrulcicek]

Dear Paulo, Thanks for fixing the issue very quickly. I'd like to ask another question related to problem definition. When running the code, if we decide to visit a particular node in an instance and the tour time exceeds the closing time of this node at the time of visit, i.e. after random noise is added, 'self.feas' variable is immediately set to 'False', and there is no way to set it back to 'True', i.e., decision causes an infeasibility for the instance. Is this due to problem definition? Imho, the only infeasibility occurs if the tour time exceeds the maximum tour length, i.e. self.maxT, because, even though a particular node doesn't yield a reward when visited after its closing time, another node which would be visited after traversing this particular node would still have a chance to yield reward unless the maximum tour length is not exceeded. I'd appreciate it if you clarify this issue for me. Thanks again. Cihan Paulo da Costa ***@***.***>, 14 May 2021 Cum, 14:12 tarihinde şunu yazdı:

…

Hi @cihantugrulcicek <https://github.com/cihantugrulcicek>, Thanks for bringing this to our attention. I have updated the code to read from_file=True when using the ''batch_env_rl''. An example of how you can call it can be found below. Note that when reading files with a different number of nodes, the instance features are padded with zeros when the number of nodes is smaller than the maximum number of nodes in the files. import os base_dir = os.path.dirname(os.getcwd()) x_dir = os.path.join(base_dir, 'data', 'valid', 'instances') adj_dir = os.path.join(base_dir, 'data', 'valid', 'adjs') batch_env = BatchEnvRL(n_envs=2, from_file=True, x_dir=x_dir, adj_dir=adj_dir) Please let me know if this is working now. Thanks again for your feedback! — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#1 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ALFWJAXI4E2F6RO6SEJYLQTTNUATFANCNFSM44ZOSWDQ> .

[paulorocosta]

Hi @cihantugrulcicek,

Yes, our interpretation is that if you violate one of the time windows from above (you can still be early without any penalty) we deem the solution 'infeasible'.
This is also true for when you exceed maxT.
Note that if you only exceed the time windows a couple of times there is still a chance you can get a 'good' tour out of it even though it's 'infeasible'.
I hope this clarifies the matter.

Cheers,

Paulo