Gianfilippo Coppola

thanks. you are right, clearly got confused.....I fixed it. The VCF is ok. I just sent you a few lines. Now, I rerun it, but got another error at preprocess...

Hi, I generated the truth.vcf with BAMSurgeon and the reference file is then same

Hi, thanks for the help! I can now see from the file that your guess was correct. I found some bases in lower case. Below is the error (part of...

the preprocess step now works, but the train step fails with the error below. What do you think ? INFO 2020-04-06 17:24:37,805 make_weights_for_balanced_classes count length classes: [(0, 3846), (1, 1753),...

I am using the --ensemble already (see below the command line). python $NEUSOMATIC/train.py --candidates_tsv $DIR/dataset/*/candidates*.tsv --out $DIR --num_threads $Ncores --batch_size 100 --ensemble

yes

below is the preprocess call python $NEUSOMATIC/preprocess.py --mode train --reference hg38.fa --region_bed $INTERVALFILE --tumor_bam syntheticTumor.bam --normal_bam syntheticNormal.bam --work $DIR --ensemble_tsv ensemble_ann.tsv --truth_vcf synthetic.vcf.gz --min_mapq 10 --num_threads 20 --scan_alignments_binary $NEUSOMATIC/neusomatic/bin/scan_alignments

the zipped files (few lines) are attached. For the candidates.tsv I attached the lines from the work.0 dir. I had a typo in the command line, now corrected [Archive.zip](https://github.com/bioinform/neusomatic/files/4441178/Archive.zip)

it is working now, going through the iterations. I will let you know if it completes. Thanks!

the network completed training. Below are the last few output lines INFO 2020-04-07 01:44:34,805 train_neusomatic epoch: 999, iter: 57925, lr: 0.0001, loss: 0.20255 INFO 2020-04-07 01:44:39,472 train_neusomatic epoch: 999, iter:...