Error when starting docker container with docker-compose: failed to create shim task: OCI runtime create failed

I have a certain code developed using Docker, and Docker-compose. When I want to start the images that I have build using:

sudo docker-compose -f docker-services-dev.yml up -d

I get the following error:

/home/ubuntu/.local/lib/python3.6/site-packages/paramiko/transport.py:33: CryptographyDeprecationWarning: Python 3.6 is no longer supported by the Python core team. Therefore, support for it is deprecated in cryptography and will be removed in a future release.
  from cryptography.hazmat.backends import default_backend
WARNING: Found orphan containers (c4kafka_zookeeper_1, c4kafka_kafka_1) for this project. If you removed or renamed this service in your compose file, you can run this command with the --remove-orphans flag to clean it up.
Recreating c4kafka_capture_1   ... done
Recreating c4kafka_detection_1 ... error
Recreating c4kafka_sender_1    ... 
Recreating c4kafka_tracking_1  ... 

ERROR: for c4kafka_detection_1  Cannot start service detection: failed to create shim task: OCI runtime create failed: runc create failed: unable to start container process: error during container init: error running hook #0: error running hook: exit status 1, stdout: , stderr: exec command: [/usr/bin/nvidia-container-cli --load-kmods configure --ldconfig=@/sbin/ldconfig.real --device=all --compute --compat32 --graphics --utility --video --display --pid=1998 /var/lib/docker/overlay2/81241e5a3cbe27712710e1bc3621781c9419021e2a7b42e4f45852d2fafc6801/merged]
Recreating c4kafka_sender_1    ... done
Recreating c4kafka_tracking_1  ... done

ERROR: for detection  Cannot start service detection: failed to create shim task: OCI runtime create failed: runc create failed: unable to start container process: error during container init: error running hook #0: error running hook: exit status 1, stdout: , stderr: exec command: [/usr/bin/nvidia-container-cli --load-kmods configure --ldconfig=@/sbin/ldconfig.real --device=all --compute --compat32 --graphics --utility --video --display --pid=1998 /var/lib/docker/overlay2/81241e5a3cbe27712710e1bc3621781c9419021e2a7b42e4f45852d2fafc6801/merged]
nvidia-container-cli: mount error: file creation failed: /var/lib/docker/overlay2/81241e5a3cbe27712710e1bc3621781c9419021e2a7b42e4f45852d2fafc6801/merged/usr/lib/libvisionworks.so: file exists: unknown
ERROR: Encountered errors while bringing up the project.

The docker-services-dev.yml file is the following:

version: '3.8'

services:

  # Services of each module needed for the project.
  capture:
    build:
      context: ../c4capture
      dockerfile: Dockerfile
      # Name of the json file with module configuration
      args:
          JSON_FILE: capture
          UTILS_VERSION: "latest"
    volumes:
      - ../images:/home/images
    restart: unless-stopped
    # Secrets wanted to pass to the module
    secrets:
      - camera_config-username
      - camera_config-password
      - camera_config-url

  detection:
    build:
      context: ../c4detection
      dockerfile: Dockerfile
      # Name of the json file with module configuration
      args:
        JSON_FILE: detection
        UTILS_VERSION: "latest"
    # Path of the neural network and name of the weights folder
    volumes:
      - ../images:/home/images
    restart: unless-stopped
    deploy:
      resources:
        reservations:
          devices:
          - driver: nvidia
            count: all
            capabilities: [ gpu ]
  tracking:
    build:
      context: ../c4tracking
      dockerfile: Dockerfile
      # Name of the json file with module configuration
      args:
        JSON_FILE: tracking
        SHAPES_FILE: shapes
        UTILS_VERSION: "latest"
    # Path of the neural network and name of the weights folder
    volumes:
      - ../images:/home/images
    restart: unless-stopped

  sender:
    build:
      context: ../c4sender
      dockerfile: Dockerfile
      # Name of the json file with module configuration
      args:
        JSON_FILE: sender
        UTILS_VERSION: "latest"
    volumes:
      - ../images:/home/images
    restart: unless-stopped
    # Secrets wanted to pass to the module
    secrets:
      - database_config-url
      - database_config-token
      - database_config-token2




# Set all service secrets and their file location
secrets:
  camera_config-url:
    file: ./secrets/camera_config-url.txt
  camera_config-username:
    file: ./secrets/camera_config-username.txt
  camera_config-password:
    file: ./secrets/camera_config-password.txt


#  ftp_config-ip:
#    file: ./secrets/ftp_config-ip.txt
#  ftp_config-username:
#    file: ./secrets/ftp_config-username.txt
#  ftp_config-password:
#    file: ./secrets/ftp_config-password.txt

  database_config-url:
    file: ./secrets/database_config-url.txt
  database_config-token:
    file: ./secrets/database_config-token.txt
  database_config-token2:
    file: ./secrets/database_config-token2.txt

If I comment the lines of

deploy:
      resources:
        reservations:
          devices:
          - driver: nvidia
            count: all
            capabilities: [ gpu ]

in the c4Detection module, then I can build all the containers, but c4Detection docker still crashes due to libcurand.so.10. I think that the error is because I have L4T-R32.3.1. installed on the Jetson TX2, and to run YOLOv5 we need PyTorch 1.8 which can't be installed. Is this the reason?