Training Tiberius and containers
agesantos opened this issue · 6 comments
Hi,
Great work!
Do you think that, if trained, Tiberius will provide equally good results for other groups of organisms other than mammals? Such as in large invertebrate genomes?
Also, do you plan to make a docker or singularity containers available for the software?
Installation can get quite complicated when your admin restricts your permission, such as in institutional Clusters, and having a container would make it much easier.
Thank you,
andré
Hi andré,
thank you for your interest.
At the moment, we have trained our model only on mammals. However, we plan to train the model on other clades aswell. We expect that the training process will be similarly effective for these groups and could also be done by others.
We may add a Docker/Singularity container in the coming weeks or months, as time allows. In the meantime, you might consider installing TensorFlow and the other required packages using Conda (see instructions here and here).
Best,
Lars
Hi andré, thank you for your interest.
At the moment, we have trained our model only on mammals. However, we plan to train the model on other clades aswell. We expect that the training process will be similarly effective for these groups and could also be done by others.
We may add a Docker/Singularity container in the coming weeks or months, as time allows. In the meantime, you might consider installing TensorFlow and the other required packages using Conda (see instructions here and here).
Best, Lars
hello,I follow this step,but i can not find tiberius.py,do you know why ?thanks
tiberius
Any progress in "train the model on other clades" ?
I will delete the out of office responses to avoid confusion ;-)
Training Tiberius is not really a problem. Selecting good training data, on the other hand, can in some cases be tricky. We got funding to develop an approach for training data selection for clades. We will start working on it around January 2025.
Hope all models coming soon.
OK，thanks