Skip to content

FAQ

Python

Q: How do I install my own Python packages?
A: See our page here

Nek5000

Q: Why does Nek5000 fail with an error "Vanishing jacobian"?
A: There is a bug in older versions of Nek5000, which has been fixed with this patch. Please pull the latest Nek5000 from Git.

Guppy

Q: Why does guppy_barcoder fail to run on GPU nodes when the --detect_mid_strand_barcodes option is used?
     Error: "Exception thrown in Barcoder worker thread: Too many sequences for GPU threads to contain! Will continue with next input file."
A: This is a known bug in guppy_barcoder. To workaround the issue, explicitly specify the barcoding kit names (--barcode_kits) along with the --detect_mid_strand_barcodes option.
      eg: guppy_barcoder -i /data/guppy/abc -s out/ --trim_barcodes --detect_mid_strand_barcodes -x auto --barcode_kits "SQK-RBK004"

SRA Toolkit

Q: Why do I get disk quota exceeded messages when using SRA Toolkit (e.g. prefetch)? I have lots of room left in my project area.
A: By default, SRA Toolkit downloads files to a cache in your home directory (/home/username/ncbi/public/sra). To change this, follow this guide

Tensorflow GPU

Q: If I install Tensorflow GPU using Pip, what modules do I need to load from the module system to make it work?
A: Tensorflow requires CUDA and CUDnn, as well as the installed pip package, to work. You can see which CUDA and CUDnn a Tensorflow Pip release requires here