adding pytorch DDP script of detection task

[sarjil77]

in this script i have implemented the DDP training for detection task.

[felixdittrich92]

Hi @sarjil77 👋,

Thanks for your PR i left some comments here :D

Before you push your code please run make style to fix the formatting 👍

If all changes are applied it would be create if you could also update the corresponding readme please ?

File: https://github.com/mindee/doctr/blob/main/references/detection/README.md

content to add:

Multi-GPU support (PyTorch only - Experimental)

Multi-GPU support on recognition task with PyTorch has been added. It'll be probably added for other tasks.
Arguments are the same than the ones from single GPU, except:

• --devices: by default, if you do not pass --devices, it will use all GPUs on your computer.
  You can use specific GPUs by passing a list of ids (ex: 0 1 2). To find them, you can use the following snippet:

import torch
devices = [torch.cuda.device(i) for i in range(torch.cuda.device_count())]
device_names = [torch.cuda.get_device_name(d) for d in devices]

• --backend: you can specify another backend for DistribuedDataParallel if the default one is not available on
  your operating system. Fastest one is nccl according to PyTorch Documentation.

python references/detection/train_pytorch_ddp.py db_resnet50 --train_path path/to/your/train_set --val_path path/to/your/val_set --epochs 5 --devices 0 1 --backend nccl

Same as you can find here:

https://github.com/mindee/doctr/blob/main/references/recognition/README.md

[felixdittrich92]

import multiprocessing as mp

That's required for the worker count if not directly passed

[felixdittrich92]

->

"""
Args:
----
    rank (int): device id to put the model on
    world_size (int): number of processes participating in the job
    args: other arguments passed through the CLI
"""

[sarjil77]

all the changes as you suggested are done along with its updated README file as well

[felixdittrich92]

Hi @sarjil77 👋

Thanks for the updates only a few small things left and we are good to merge 👍

After changes are applied please run:

make style

or

pip3 install ruff
ruff format .
ruff check --fix .

[felixdittrich92]

missing import

import multiprocessing as mp

[felixdittrich92]

Please remove the

----

-->

"""
Args:
    rank (int): device id to put the model on
    world_size (int): number of processes participating in the job
    args: other arguments passed through the CLI
"""

We changed this in the meanwhile :)

[felixdittrich92]

@sarjil77 if rank == 0 missing :)

[sarjil77]

i have made all the changes a s required :), and i am not adding "import multiprocessing as mp" as mp bacause mp is already defined using "pytorch.multiprocessing as mp", i have taken this reference from DDP script of Recognition task and i think it is good to merge now. :)

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 96.55%. Comparing base (83f1bc5) to head (a522f0e).
Report is 10 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1777      +/-   ##
==========================================
- Coverage   96.56%   96.55%   -0.02%     
==========================================
  Files         164      165       +1     
  Lines        7895     7901       +6     
==========================================
+ Hits         7624     7629       +5     
- Misses        271      272       +1     

Flag	Coverage Δ
unittests	96.55% <ø> (-0.02%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

---

🚨 Try these New Features:

• Flaky Tests Detection - Detect and resolve failed and flaky tests

[felixdittrich92]

Hi @sarjil77 👋,

Thanks for the updates 👍

2 small things left but don't worry i can do this thanks for the PR :)