Unsupported CIGAR operation when computing inclusive scan #83
Description
Hi,
I am trying to correct about 40x of HG002 long reads using the experimental R9.4 model. Admittedly, I am using Herro as implemented in dorado 1.1 (dorado correct) for which I have swapped the R10 model for the R9 model. I have first generated my alignment file which resulted in a ~900GB PAF file. I have then run the correction of my reads using that PAF file using a DGX A100.
About 70% of the correction in (3.6 million reads corrected, 1.5 million left to correct), the run is interrupted by the following error:
terminate called after throwing an instance of 'std::runtime_error'
what(): Unsupported CIGAR operation when computing inclusive scan!
There is unfortunately no indication of which record is the problem. I have tried to resume the run but ended up on the same issue. I have tried to restarting the run but it also got interrupted with the exact same quantity of reads corrected in output.
I have also tried to see if I could skip the problematic record: I computed the list of 1.5 million uncorrected reads (in the order they appear in the original fastq file), split that list it in 2 (let's call those part1 and part2), and run 2 corrections where I told dorado to resume the correction from my list of previously corrected reads AND either part1 or part2. For one of the 2 runs, I ended up with the same interruption and error message as above. For the other one, dorado completed successfully BUT the log says that Number of reads submitted for correction: 25469, which is clearly incorrected as there were 700,000+ reads left to correct, and the output is empty.
I would appreciate any help on this issue.
Thanks!
Guillaume