IQTree_Crashes: Two_DATASETS_TWO_ERRORS

[Tarun JaiRaj Narwani]

Dear Devs@IQTree

I am trying to run IQTREE_v_1.4.2 on two datasets of 1421 Seq and 118 Sequences. Both are the Multiple Sequence Alignment files.
IQtree crashes in both cases but with different errors.

IQtree_1421:
Command: iqtree-omp -st AA -nt 16 -wbtl -b 1000 -abayes -con -m TESTNEW -madd C20,C60,EX2,EX3,EH0,UL2,UL3,EX_EHO,LG4M,LG4X -s MSA_118.phy
Error: ERROR: Not enough memory, allocation of 49476787216 bytes failed (bad_alloc)  (For details checkout the attached Log File_ outStd_1421)


IQtree_118:
Command: iqtree-omp -st AA -nt 16 -wbtl -b 1000 -abayes -con -m TESTNEW -madd C20,C60,EX2,EX3,EH0,UL2,UL3,EX_EHO,LG4M,LG4X -s MSA_118.phy
Error: ERROR: Invalid model name EH0. (Log File: outStd_118)

Why does the first one throws a memory allocation (RAM) error, when I have enough space in my cluster's RAM?
Why does the error of Invalid model name does not appear in IQTree1421?

[Bui Quang Minh]

Hi Tarun,

On May 17, 2016, at 5:30 PM, Tarun JaiRaj Narwani <tjrna...@gmail.com> wrote:

Dear Devs@IQTree

I am trying to run IQTREE_v_1.4.2 on two datasets of 1421 Seq and 118 Sequences. Both are the Multiple Sequence Alignment files.
IQtree crashes in both cases but with different errors.

IQtree_1421:
Command: iqtree-omp -st AA -nt 16 -wbtl -b 1000 -abayes -con -m TESTNEW -madd C20,C60,EX2,EX3,EH0,UL2,UL3,EX_EHO,LG4M,LG4X -s MSA_118.phy
Error: ERROR: Not enough memory, allocation of 49476787216 bytes failed (bad_alloc)  (For details checkout the attached Log File_ outStd_1421)

the error message says that IQ-TREE requires ~49GB, whereas you computer has only ~31 GB as reported at the beginning of the run:

Host:    r1i0n4 (AVX, 31 GB RAM)

For your more information, this is the RAM required for the C60 model. If you don’t have computer with that much memory, remove C60 from the list. Otherwise (if you have another high-memory PC), then copy the checkpoint file .ckp.gz and the model file (.model) into this PC, there you can resume the analysis using the same command line, so that no computation time is lost.

IQtree_118:
Command: iqtree-omp -st AA -nt 16 -wbtl -b 1000 -abayes -con -m TESTNEW -madd C20,C60,EX2,EX3,EH0,UL2,UL3,EX_EHO,LG4M,LG4X -s MSA_118.phy
Error: ERROR: Invalid model name EH0. (Log File: outStd_118)

Why does the first one throws a memory allocation (RAM) error, when I have enough space in my cluster's RAM?

see above

Why does the error of Invalid model name does not appear in IQTree1421?

There is a typo, this model name should be EHO (letter O) instead of EH0 (zero) in your command line.

Cheers, Minh

--
You received this message because you are subscribed to the Google Groups "IQTREE" group.
To unsubscribe from this group and stop receiving emails from it, send an email to iqtree+un...@googlegroups.com.
To post to this group, send email to iqt...@googlegroups.com.
Visit this group at https://groups.google.com/group/iqtree.
For more options, visit https://groups.google.com/d/optout.
<outStd_1421><outStd_118>

--

Bui Quang Minh

Center for Integrative Bioinformatics Vienna (CIBIV)

Campus Vienna Biocenter 5, VBC5, Ebene 1

A-1030 Vienna, Austria

Phone: ++43 1 4277 74326

Email: minh.bui (AT) univie.ac.at

[Tarun JaiRaj Narwani]

Dear Minh

Please find my replies inline.

On May 17, 2016, at 5:30 PM, Tarun JaiRaj Narwani <tjrna...@gmail.com> wrote:

Dear Devs@IQTree

I am trying to run IQTREE_v_1.4.2 on two datasets of 1421 Seq and 118 Sequences. Both are the Multiple Sequence Alignment files.
IQtree crashes in both cases but with different errors.

IQtree_1421:
Command: iqtree-omp -st AA -nt 16 -wbtl -b 1000 -abayes -con -m TESTNEW -madd C20,C60,EX2,EX3,EH0,UL2,UL3,EX_EHO,LG4M,LG4X -s MSA_118.phy
Error: ERROR: Not enough memory, allocation of 49476787216 bytes failed (bad_alloc)  (For details checkout the attached Log File_ outStd_1421)

the error message says that IQ-TREE requires ~49GB, whereas you computer has only ~31 GB as reported at the beginning of the run:

Host:    r1i0n4 (AVX, 31 GB RAM)

For your more information, this is the RAM required for the C60 model. If you don’t have computer with that much memory, remove C60 from the list. Otherwise (if you have another high-memory PC), then copy the checkpoint file .ckp.gz and the model file (.model) into this PC, there you can resume the analysis using the same command line, so that no computation time is lost.

Yes, I just confirm that RAM per node is 31 GB. I am running IQtree on a cluster of 48 computers each with similar no. of RAM, therefore I can access far much greater RAM only iff I can use MPI. I searched for IQTree MPI version and could not find it.
Can you direct me towards a mpi version of IQtree, if available?

IQtree_118:
Command: iqtree-omp -st AA -nt 16 -wbtl -b 1000 -abayes -con -m TESTNEW -madd C20,C60,EX2,EX3,EH0,UL2,UL3,EX_EHO,LG4M,LG4X -s MSA_118.phy
Error: ERROR: Invalid model name EH0. (Log File: outStd_118)

Why does the first one throws a memory allocation (RAM) error, when I have enough space in my cluster's RAM?

see above

Why does the error of Invalid model name does not appear in IQTree1421?

There is a typo, this model name should be EHO (letter O) instead of EH0 (zero) in your command line.

I apologize for this one.. My Bad..
Thanks..!! :)

Cheers, Minh

[Tung Nguyen]

Dear Tarun, 

The MPI version of IQ-TREE is not quite ready yet because we still need to test it thoroughly. Nevertheless, even if you use the MPI version your problem with the memory limitation will not be solved. In fact, we do not distribute the memory but rather the computation. Thus, IQ-TREE-MPI will still require the same amount of memory for each node. 

Distributing the memory over different cluster nodes will make the computation very slow. In every analysis, the tree likelihood is computed millions of times. If the memory is distributed, then every time the tree likelihood is computed the cluster nodes need to "talk" to each other to collect the results. Thus, the performance will heavily depend on your network speed and the number of nodes you can use will be limited. 

I would suggest that you do what Minh already suggested: remove the C60 from the model list to reduce memory consumption, C60 is rather an overkill. 

Cheers

Tung  

--