I’m not being able to create a taxonomic classifier for SILVA 128, 97% or 99% OTUs. I think I run out of memory.
In Data Resources webpage, there is a link for downloading Silva 119 99% OTUs full-length sequences, but this version was trained using a previous version of scikit-learn 0.19.0 and thus no longer works with the current versions of Qiime2 (this is the message that comes up in my computer when I try to use this file).
Would it be possible to provide a link to download the full-length sequences for SILVA 128 database, either 97% or 99% OTUs, that would be compatible with the version of scikit-learn/Qiime2?
This is certainly consistent with what we have seen, regarding memory usage while training with the Silva database. There isn’t much else to do besides utilize a computer with additional memory resources. You can use one of the QIIME 2 AWS machines, if you don’t have access locally to something bigger.
Hmm, something isn’t quite right here — I just performed a fresh install of QIIME 2 2017.10 (the latest and greatest Q2) and pulled down the Silva 119 99% OTUs full-length sequences (built against sklean 0.19.0), and this appears to work just fine (e.g. I am able to classify features without error or warning).
Are you running a development version of QIIME 2? Or perhaps an old version of QIIME 2? Maybe you installed a newer version of sklearn in your qiime 2 conda environment? Either way, please double check your installed version of QIIME 2 (maybe just reinstall the env from scratch), and the download for the latest version of the silva classifier (the MD5 sum is listed on the docs site, so that you can be absolutely sure you grabbed the right file).