Hi, I'Ve Been Trying To Use Hyperparameter Optimization For Yolov11. When I Try To Get The Top Metrics Using

Answered

Hi,
I've been trying to use hyperparameter optimization for Yolov11. When I try to get the top metrics using

# optimization is completed, print the top performing experiments id
k = 3
top_exp = optimizer.get_top_experiments(top_k=k)
print('Top {} experiments are:'.format(k))
for n, t in enumerate(top_exp, 1):
    print('Rank {}: task id={} |result={}'
          .format(n, t.id, t.get_last_scalar_metrics()['val']['metrics/mAP50(B)']['last']))

t.get_last_scalar_metrics() returns an empty dict. I've tried using different titles and series names but haven't had any success. Any ideas on what I'm missing here?
Thanks!

  				
Posted 
	2 months ago

					More  		
  Report
		
					ContemplativeParrot88
				
					0
					 × 1

Votes Newest

Answers 6

Hi CostlyOstrich36 I ran the example provided here:
None
and I can see the hyperparameters and the reported scalars for the base task. Running the optimizer doesn't show any errors but printing the results from the top experiments returns this:

Rank 1: task id=0b9a695218424ac7bb42eb7cca490f72 |result={}
Rank 2: task id=a599229be44a48f0b77e3cad2e6bd4a4 |result={}
Rank 3: task id=b70bebfa88d54e6c96b73ce22c95c405 |result={}

  				
Posted 
	2 months ago

					More  		
  Report
		
					ContemplativeParrot88
				
					0
					 × 1

Hi ContemplativeParrot88 , are you sure the hyper parameters themselves are properly connected? If you run as a single run and change parameters, do they take effect?

  				
Posted 
	2 months ago

					More  		
  Report
		
					CostlyOstrich36
				
					0

Hi ContemplativeParrot88 ! Are the scalars in the UI in the optimization tasks (not the base task)?

  				
Posted 
	2 months ago

					More  		
  Report
		
					SmugDolphin23
				
					0

The scalars are in the base task. They don't show up in the optimization task

  				
Posted 
	2 months ago

					More  		
  Report
		
					ContemplativeParrot88
				
					0
					 × 1

Then there is likely a problem with those tasks. For example, could be that the hyper parameters get values that are too low or high which just bugs out the training.

  				
Posted 
	2 months ago

					More  		
  Report
		
					SmugDolphin23
				
					0

CostlyOstrich36 Any ideas on how I can resolve this?

  				
Posted 
	2 months ago

					More  		
  Report
		
					ContemplativeParrot88
				
					0
					 × 1

Write your answer

284 Views

6 Answers

2 months ago