search

lbcb-sci / raven

De novo genome assembler for long uncorrected reads
MIT License
202 stars 21 forks source link

QUESTION: CUDA Parameters Optimisation? #41

Closed cement-head closed 3 years ago

cement-head commented 3 years ago

Trying to optimise the CUDA parameters. Using 24 GB GPUs (2x RTX TITANs).

This is essentially the command I'm using, but not sure if I'm getting proper CUDA usage...

$ raven -t 124 -c 100 -a 100 input_file.fastq > raven_asm.fasta

Screenshot from 2021-03-02 09-37-12

cement-head commented 3 years ago

For RACON:

only available when built with CUDA:
    -c, --cudapoa-batches <int>
        default: 0
        number of batches for CUDA accelerated polishing per GPU
    -b, --cuda-banded-alignment
        use banding approximation for polishing on GPU. Only applicable when -c is used.
    --cudaaligner-batches <int>
        default: 0
        number of batches for CUDA accelerated alignment per GPU
    --cudaaligner-band-width <int>
        default: 0
        Band width for cuda alignment. Must be >= 0. Non-zero allows user defined
        band width, whereas 0 implies auto band width determination.

For RAVEN: 

only available when built with CUDA:
    -c, --cuda-poa-batches <int>
      default: 0
      number of batches for CUDA accelerated polishing
    -b, --cuda-banded-alignment
      use banding approximation for polishing on GPU
      (only applicable when -c is used)
    -a, --cuda-alignment-batches <int>
      default: 0
      number of batches for CUDA accelerated alignment

Is there some detailed explanation on how to optimise and/or use these parameters for CUDA?

rvaser commented 3 years ago

The parameters are the same in Racon and Raven, there are small differences in names and band width is missing in Raven. Mapping in Racon/Raven is not yet on the GPU, only alignment and POA in Racon are and hence the nvidia-smi does not show any GPU usage yet.

@tijyojwad, can you please assist @cement-head with parameter tuning for cuda-poa/cuda-aligner parameters?

P.s. Sorry for my late reply!

cement-head commented 3 years ago

If the RAVEN genome assembly is too large 7 Gbp rather than the expected 6 Gbp, would the best place to focus be on (a) the polishing step, or (b) the initial assembly parameters step?

msikic commented 3 years ago

have you tried to purge haplotigs with purge_dups https://github.com/dfguan/purge_dups

On Wed, Mar 10, 2021 at 10:21 PM Andor J Kiss @.***> wrote:

If the RAVEN genome assembly is too large 7 Gbp rather than the expected 6 Gbp, would the best place to focus be on (a) the polishing step, or (b) the initial assembly parameters step?

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/lbcb-sci/raven/issues/41#issuecomment-795479586, or unsubscribe https://github.com/notifications/unsubscribe-auth/AALV5JAPV2SEDFBFY7F3Q3LTC5W7BANCNFSM4YPSDJAA .

cement-head commented 3 years ago

Nope. I haven't.