ERROR building database for HHsuite using cstranslate

[bioxruan]

Hello,
I am trying to build a database for HHblits. I have the database was first clustered by mmseqs, and customed for hhsuite.

I use the command as below:

mmseqs cluster DB clu tmp
mmseqs result2msa DB DB clu cluMsa --compressed 0
ln -s DB_h cluMsa_header.ffdata
ln -s DB_h.index cluMsa_header.ffindex
ln -s DB cluMsa_sequence.ffdata
ln -s DB.index cluMsa_sequence.ffindex

and then run:
cstranslate -i cluMsa -o cluMsa_cs219 -x 0.3 -c 4 -I ca3m

I got error:
Reading context library for pseudocounts from internal ...
Reading abstract state alphabet from internal ...

ERROR: Sequence 1 has 3446 match columns but should have 1847!

How can I solve it.

I am using:
MMseqs2 Version: 14-7e284
HHsuite 3.3.0

Thanks