Here's a refined version of the README file for the project:

This project is built upon the CodaLab Competitions framework and has been customized to utilize PSC as the evaluation server. The following diagram illustrates the workflow:

The project is deployed at https://ears.urgent-challenge.com. An administrator account has been created for access:

• Username: urgent
• Password: g3!yD3CWNbf7gzdu

You can visit the competition page directly at: Competition Page

For more detailed information, refer to the CodaLab Documentation.

To access the urgent server, follow these steps:

1. Copy the file examples/urgent.pem to ~/.ssh/urgent.pem
2. Add the following configuration to your ~/.ssh/config file:

   Host urgent
       HostName 52.26.172.172
       User ubuntu
       Port 22
       IdentityFile ~/.ssh/urgent.pem

3. The urgent EARS project is located at /home/ubuntu/workspace/urgent-ears.

To create a new competition, you need to set up a new evaluation task queue and a Celery worker.

1. Go to Worker Queue Management and create a new queue.
2. Copy the broker URL, add a new worker to docker-compose.yml, and run docker compose up -d. Below is an example configuration for a new worker. Replace 'ears.urgent-challenge.com' in the broker URL with 'rabbit' since we will only access the broker within the Docker network.
3. Update the TRACK parameter to match the track used by the PSC evaluation program.
4. Prepare the evaluation dataset on PSC
5. Upload the competition bundle for the new competition. An example bundle is available at examples/competition.zip.

worker_ears:
  build:
    context: worker
    dockerfile: Dockerfile
  restart: unless-stopped
  env_file: .env
  environment:
    - TRACK=ears
    - BROKER_URL=pyamqp://7f541161-6af6-43a3-821c-4332006d7a97:40754acf-6723-49b9-b81f-ef06fa7b5106@rabbit:5672/394337fa-550b-47e3-bcf5-2c32e4e14baa
  hostname: worker_ears
  links:
    - rabbit
  logging:
    options:
      max-size: "200k"
  mem_limit: 1g
  memswap_limit: 1g
  container_name: worker-ears

For the evaluation program, refer to /ocean/projects/cis210027p/wwang13/workspace/urgent/evaluation. Prepare a directory named workspace/<track>/assets. Refer to the EARS track as an example.

workspace/ears/assets/
|-- models
|   |-- model_v8.onnx -> /ocean/projects/cis210027p/wzhangn/espnet_urgent2024/egs2/urgent24/enh1/urgent2024_challenge/DNSMOS/DNSMOS/model_v8.onnx
|   `-- sig_bak_ovr.onnx -> /ocean/projects/cis210027p/wzhangn/espnet_urgent2024/egs2/urgent24/enh1/urgent2024_challenge/DNSMOS/DNSMOS/sig_bak_ovr.onnx
`-- reference
    |-- clean
    `-- spk1.scp