wandreopoulos
commented
2 months ago Hi, Sorry for the delay. I haven't been monitoring this email account regularly. I would suggest making a new branch, and then making a pull request, if you wish to do secondary development.
You may retrain the model with the train script (no Docker build is needed for re-training, since your main goal is the model with the h5 files). The information for training is in the Supplementary Information document. The training files I used are on the NERSC link in the README file.
I recently also pushed a new Docker container.
To run for example: ~/Downloads/deeplasmid/classifier/dl$ sudo
/usr/bin/docker run -it -v
pwd/testing/649989979/649989979.fna:/srv/jgi-ml/classifier/dl/in.fasta -v
pwd/testing/649989979/649989979.fna.OUT:/srv/jgi-ml/classifier/dl/outdir
billandreo/deeplasmid.tf.gpu3 deeplasmid.sh in.fasta outdir
I also revamped the README to explain how to run and rebuild the Docker container. A few tools and files are needed to rebuild the Docker image, as described in the README. Also these will be needed if you plan to retrain the model:
These are some things to consider if re-building the Docker image (also described in the Supplementary Information from the publication). These need to be available under the top build directory (they will be copied in the Docker container): - Prodigal, use git clone https://github.com/hyattpd/Prodigal.git - hmmer (v3.3.2 was used initially)
- bbtools/sketch (BBMap_38.73.tar.gz was used initially) The models, sketch and pfam files can be downloaded from https://portal.nersc.gov/dna/microbial/assembly/deeplasmid/ : - the model .h5 files from training should be under Plasmid_Models/plasmid4z-newfeat12-?? subdirs. These can be downloaded quickly from nersc with wget - several sketch files should be under the asafl_plasmidPred subdirectory - the files pfams_discr.txt and Pfam-A.TMP2.hmm under the top build directory
On Tue, Jan 9, 2024 at 12:43 AM a-piece-of-teemo @.***> wrote:
Hello author, Does your model support secondary development? If I want to make changes in your code, how should I retrain the model and what commands should I use to run the code? Looking forward to your reply.
— Reply to this email directly, view it on GitHub https://github.com/wandreopoulos/deeplasmid/issues/12, or unsubscribe https://github.com/notifications/unsubscribe-auth/AANGW5L7LQ6KVJEMY2QHNFTYNT7KFAVCNFSM6AAAAABBSW4522VHI2DSMVQWIX3LMV43ASLTON2WKOZSGA3TCOBZGAZTGOA . You are receiving this because you are subscribed to this thread.Message ID: @.***>
-- Thanks, Bill
William B. Andreopoulos, Ph.D. Joint Genome Institute LBNL
Hello author, Does your model support secondary development? If I want to make changes in your code, how should I retrain the model and what commands should I use to run the code? Looking forward to your reply.