Training using the commands given by the repository cannot converge on the biycle dataset

[guagua11]

Thank you very much for your great work！
I use the training command recommended by the repository to train on the public bicycle dataset and can't converge, please help to see where the problem is, thank you! The command to run is as follows:

python train.py -s /home/xx/lab/SteepGS/data/bicycle -m /home/xx/lab/SteepGS/output/bicycle --no_gui --densify_strategy steepest --S_estimator partial --eval

even i used
python train.py -s /home/xx/lab/SteepGS/data/bicycle -m /home/xx/lab/SteepGS/output/bicycle --no_gui --densify_strategy steepest --S_estimator approx --eval
and
python train.py -s /home/xx/lab/SteepGS/data/bicycle -m /home/xx/lab/SteepGS/output/bicycle --no_gui --densify_strategy steepest --S_estimator inv_cov --eval

None of them could converge！

[zhj296409022]

Same here.

[loryruta]

Same here

[loryruta]

Please, could you share the exact hyperparameters to reproduce the paper results?

[Aristomir]

Same here. Fails to converge on multiple datasets

[FlyPeo]

Same here