Not parsing flags from toml file

[adamhorvath99]

Dear all,

I might be wrong, but I have just updated to the 3.0.2 version of crest, and somehow the parsing of flags arguments from toml file is misteriously not working now. I use the --cinp flag but it is uneffective.

---

• INPUT FILE ./TS3_difoszfinin_search4.toml content (without comments):

  ---

• input = "geom.xyz"
• threads = 28
• [calculation]
• type = 1
• eprint = true
• elog = "energies.log"
• [[calculation.level]]
• method = "gfn2"
• binary = "/scratch/c_radstab/conformer/xtb-dist"
• uhf = 0
• chrg = 0

• flags = "--cinp constraints.inp --gfn2 --grad"

  ---

  ---

  Initial Geometry Optimization

  Geometry successfully optimized.

  WARNING Change in topology detected! Topology change compared to the input affects atoms: 1(C) 17(C)

[pprcht]

You are using a weird setup that is mixing up options, and I don't think there is a bug:

• method="gfn2" tells the program to use tblite, not the xtb binary, yet
• binary=... requests an xtb binary with
• these flags flags=..., in which
• --cinp is a CREST option, so xtb doesn't know about this

If you tell me what exactly you want to do in all this, I might be able to better assist.

[adamhorvath99]

Thank you very much! I am really lost at this point now.

I want two thing. First, do usual conf-analysis with crest using gfn2. Second, I want constrained conf-analysis with crest using gfn2 for TSs. I would need TOML files fro these. I think I misunderstood a lot of things in the manual

[pprcht]

If you do not explicitly need the xtb implementation and are fine with tblite you can either use the command line as

crest geom.xyz --gfn2 --cinp constraints.inp -T 28

or a toml file

input="geom.xyz"
threads=28
constraints="constraints.inp"
[[calculation.level]]
method="gfn2"
chrg=0
uhf=0

or a combination of both

crest input.toml --cinp constraints.inp -T 28

the point being, all new versions can work without an xtb binary.

[adamhorvath99]

OK, now started to work. But I get this memory issue. I allocate 2GB per core and have a lot of scratch, I do not understand the problem.

In file '/home/runner/work/crest/crest/src/parsing/parse_xtbinput.f90', around line 720: Error allocating 2188448768777856 bytes: Cannot allocate memory

[pprcht]

This should have been fixed as of commit 5ca82feb2ec4df30a0129db957163c934f085952, are you using the current version? It might not be included in the conda build (yet)