External Email - Use Caution

Various postings reference it’s related to the application failing to obtain memory.  That does not necessarily mean virtual/physical memory is exhausted, but rather the memory management scheme used by the application is unable to provide a big enough chunk of memory to be allocated in time for use.   Some people suggest monitoring the process, e.g., “ps” command, to look at either system memory usage or GPU memory usage in the case of using GPU, e.g., NVIDIA command “nvidia-smi” while things are running.  One post seems to indicate memory is not physically underrun from a graph of memory consumption when the error occurs.

- R. https://secure-web.cisco.com/1lQSEMDdo1ulf5k6B_UJrBUyBSDW206IbKCm6lAKEaSvRWB... https://secure-web.cisco.com/167ERnV8gwZmPGQneh48CAQ4nbl0Jx8bfLYgpOKuUmdUIeN... https://secure-web.cisco.com/1PBtw1Ekc9zECMksHeg_KIzO-vNIHFECQ7w13UbgUebvkyV... https://secure-web.cisco.com/1CxLDSwk7AAkcqahJRoHD9KM8_J8VZB04y-dz5oD-weCfrJ...

On Mar 17, 2023, at 05:37, Steinbacher Jürgen <j.steinbacher@salk.at> wrote:        External Email - Use Caution        Dear Freesurfer team, I have now fs 7.3.2 installed on ubuntu 22.04.2 on a 8 core i7-4790 CPU (supporting AVX instructions)  with 28.0Gb  fs works fine now,  but I had a patient where the segmentation did not work  well, so I wanted to try  mri_synthseg.I have tensorflow 2.4.1 (python 3.7.13) installed and running on a gtx 1650 gpu. So running mri_synthseg first time did not prompt to install it again.When I run mri_synthseg  it tells predicting 1/1 and after about 25 sec it terminates telling: „terminate called after throwing an instance of ‚std::bad_alloc‘“  what(): std::bad_alloc  Abort(core dumped)?????????? (memory problem??)It terminates independent of using other patients, or gpu – cpu or more or less cores..I also tried „synthstrip“ to test the tensorflow installation working with fs and this works well!Do you have an idea, & solution??? Greetings Jürgen _______________________________________________Freesurfer mailing listFreesurfer@nmr.mgh.harvard.eduhttps://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer