Open ten-xi opened 1 month ago
24 hours for the first, and 60 hours for the next two.
On Thu, 18 Jul 2024, 09:42 ten-xi, @.***> wrote:
Dear author, I am very interested in your work, may I ask how long you run an experiment?
— Reply to this email directly, view it on GitHub https://github.com/TesfayZ/CCM_MADRL_MEC/issues/6, or unsubscribe https://github.com/notifications/unsubscribe-auth/ANZZ2746YDGFGO7GJKVXZX3ZM55YBAVCNFSM6AAAAABLCGDSRCVHI2DSMVQWIX3LMV43ASLTON2WKOZSGQYTKNZVGEZDKNY . You are receiving this because you are subscribed to this thread.Message ID: @.***>
What if you put it on cuda? Have you tried that?
Run CCM_MADRL.py for 24 hours, so run the whole experiment
I didn't. You can see the output onscreen if you don't want wait until the end. Uncomment the last lines of the ccm-madrl.py to print results at each episode.
On Thu, 18 Jul 2024, 10:35 ten-xi, @.***> wrote:
What if you put it on cuda? Have you tried that?
— Reply to this email directly, view it on GitHub https://github.com/TesfayZ/CCM_MADRL_MEC/issues/6#issuecomment-2236061605, or unsubscribe https://github.com/notifications/unsubscribe-auth/ANZZ276I3RQQARJSJDB5U63ZM6D53AVCNFSM6AAAAABLCGDSRCVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDEMZWGA3DCNRQGU . You are receiving this because you commented.Message ID: @.***>
Can you provide comments for that part of the model? checkpoint = torch.load('./checkpoint/models_checkpoint'+str(self.InfdexofResult)+'.pth')
changes = []
for agent_id in range(self.n_agents):
ce = self.check_parameter_difference(self.actors[agent_id], checkpoint['actors'][agent_id])
changes.append(ce)
# Check for parameter differences in critics
for agent_id in range(1):
ce = self.check_parameter_difference(self.critics[agent_id], checkpoint['critics'][agent_id])
changes.append(ce)
if sum(changes) >1:
#print("Model update detected", changes)
self.save_models('./checkpoint/models_checkpoint'+str(self.InfdexofResult)+'.pth')
elif sum(changes) == 1:
print("No actor model update detected", changes)
self.save_models('./checkpoint/models_checkpoint'+str(self.InfdexofResult)+'.pth')
#exit()
else:
print("No model update detected", changes)
self.save_models('./checkpoint/models_checkpoint'+str(self.InfdexofResult)+'.pth')
#exit()
'''
Is that the part?
No, go to the last line and uncomment. The code you posted was used to detect whether the weight of the neural network is updating with the training. You don't need it now.
On Thu, Jul 18, 2024 at 10:46 AM ten-xi @.***> wrote:
Can you provide comments for that part of the model? checkpoint = torch.load('./checkpoint/models_checkpoint'+str(self.InfdexofResult)+'.pth')
Check for parameter differences in actors
changes = [] for agent_id in range(self.n_agents): ce = self.check_parameter_difference(self.actors[agent_id], checkpoint['actors'][agent_id]) changes.append(ce)
Check for parameter differences in critics
for agent_id in range(1): ce = self.check_parameter_difference(self.critics[agent_id], checkpoint['critics'][agent_id]) changes.append(ce) if sum(changes) >1:
print("Model update detected", changes)
self.save_models('./checkpoint/models_checkpoint'+str(self.InfdexofResult)+'.pth') elif sum(changes) == 1: print("No actor model update detected", changes)
self.save_models('./checkpoint/models_checkpoint'+str(self.InfdexofResult)+'.pth')
exit()
else: print("No model update detected", changes)
self.save_models('./checkpoint/models_checkpoint'+str(self.InfdexofResult)+'.pth')
exit()
''' Is that the part?
— Reply to this email directly, view it on GitHub https://github.com/TesfayZ/CCM_MADRL_MEC/issues/6#issuecomment-2236084898, or unsubscribe https://github.com/notifications/unsubscribe-auth/ANZZ277Y7DPF6X77B2K2STTZM6FHTAVCNFSM6AAAAABLCGDSRCVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDEMZWGA4DIOBZHA . You are receiving this because you commented.Message ID: @.***>
Hello author, did you not consider putting these operations on the GPU at that time?If you run it on a GPU, is it much faster?
Dear author, I am very interested in your work, may I ask how long you run an experiment?