Closed benn888 closed 1 month ago
This is fairly common for virtual machines not to support long vector instructions.
On Windows you might like to try Windows Subsystem for Linux to run a Ubuntu environment. It uses Windows native hypervisor and I believe allows access to full CPU instruction sets.
WSL was indeed the solution to avoid restrictions of VM, it worked perfectly!
Thanks so much for your help!
Hi there,
I used pip to install medaka on a Virtualbox VM Ubuntu 22.04. The host machine runs on Windows 11 and has access to AVX instructions. However, when I try to run medaka_consensus, I get this error:
AVX instructions seem indeed not to be available on the Ubuntu guest machine: I get nothing when I do
grep "avx" /proc/cpuinfo
According to my readings, this is an issue with Virtualbox which does not enable AVX on the guest machine.Hence, I would like to know how to get rid of this error?
1) I am very new to this field but I think I understand that compiling a tensorflow version not requiring AVX instructions could help. Am I right? And if so, would medaka still run without AVX ? If this is a valid option, could you please provide information on how to compile tensorflow without AVX ? As I said, I am completely new to this field :)
2) Would installing medaka from source rather than from pip be a solution? I actually tried to install it that way too but I got this error message at the
make install
step:Running
sudo make install
brought me a bit further but I got another message:If installing medaka from source can help solving the no-AVX issue, what should I do to have a proper installation without the above error encountered at the
make install
step?Thank you in advance for your help!