Training time and debugging issues

[kwkw-z]

Thank you so much for the work you do, I'm now trying to run the code,
When training a Sym-Q model: Used
python train.py,
How long do you need to train with 5,000,000 data when you train? Do num_epochs have to train for 100,000 rounds?
and
Did python beam_search.py --weights_path PATH_TO_WEIGHTS --target ssdnc_r2 run into this problem when running this code?
2025-07-11-16-54-16/best.pth" --target ssdnc_r2
SymQ running on cuda:0
100%|████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 218.83it/s]
train dataset after filter has 981 examples
0%|▏ | 2/981 [14:25<117:44:01, 432.93s/it] Encountered in bfgs: '<' not supported between instances of 'complex' and 'float'
Encountered in bfgs: '<' not supported between instances of 'complex' and 'float'
Encountered in bfgs: '<' not supported between instances of 'complex' and 'float'
Encountered in bfgs: '<' not supported between instances of 'complex' and 'float'
Encountered in bfgs: '<' not supported between instances of 'complex' and 'float'
Encountered in bfgs: '<' not supported between instances of 'complex' and 'float'
Encountered in bfgs: '<' not supported between instances of 'complex' and 'float'
Encountered in bfgs: '<' not supported between instances of 'complex' and 'float'
Encountered in bfgs: '<' not supported between instances of 'complex' and 'float'
Encountered in bfgs: '<' not supported between instances of 'complex' and 'float'
Encountered in bfgs: '<' not supported between instances of 'complex' and 'float'
Encountered in bfgs: '<' not supported between instances of 'complex' and 'float'
Encountered in bfgs: '<' not supported between instances of 'complex' and 'float'
Encountered in bfgs: '<' not supported between instances of 'complex' and 'float'
Encountered in bfgs: '<' not supported between instances of 'complex' and 'float'
Encountered in bfgs: '<' not supported between instances of 'complex' and 'float'
Exception encountered: list index out of range
Encountered in bfgs: complex exponentiation
Exception encountered: maximum recursion depth exceeded while calling a Python object
Encountered in bfgs: complex exponentiation
Encountered in bfgs: complex exponentiation
1%|▋ | 8/981 [1:02:38<126:58:14, 469.78s/it]

I would appreciate it if you could reply, thank you again.

[Yuantian013]

The error message"< not supported between instances of complex and 'float'"when using scipy.optimize.minimize usually means that your objective function is returning complex numbers, but optimization routines expect real-valued (float) outputs. Most optimization methods (like 'BFGS 'SLSQP',etc.) are designed for real-valued functions and variables.

But it is just a warning, it will happen but doesn't matter

[kwkw-z]

Thank you very much for your answer, I have benefited a lot, but I have another question, that is, in training, the training epochs written in the code are 100000 epochs, this can only be trained 100 epochs a day on my device, it takes a long time, can I reduce the rounds？ Will it affect the performance of the model?I would appreciate it if you could reply, thank you again.

[Yuantian013]

300 ep can reproduce the result

[kwkw-z]

Thank you very much for your reply, I have benefited a lot！