[NEAMT] Spawn all the modules as separate services

[nikit91]

Modules like mgenre take forever to load onto the memory, which slows down the development/debugging process of the application. Having all modules as services would make the core framework lightweight and would allow faster development/debugging.

[nikit91]

For MT modules, we expect many future components to be from huggingface. In this case, we can create a docker service spawner that builds and spawns services based on a configuration. Usually, a single MT approach can have multiple models depending upon source and target language. Keeping that in mind, the configuration can look something like this:

{
  "<component_id>": [
    {
      "source_lang": "de",
      "target_lang": "en",
      "model": "<huggingface_model_id>",
      "tokenizer": "<huggingface_tokenizer_id>"
    }
  ],
  "<component_id_2>": [
    {
      "source_lang": "any",
      "target_lang": "en",
      "model": "<huggingface_model_id>",
      "tokenizer": "<huggingface_tokenizer_id>"
    }
  ],
  "<component_id_3>": [
    {
      "source_lang": "de",
      "target_lang": "en",
      "model": "<huggingface_model_id>",
      "tokenizer": "<huggingface_tokenizer_id>"
    },
    {
      "source_lang": "fr",
      "target_lang": "en",
      "model": "<huggingface_model_id>",
      "tokenizer": "<huggingface_tokenizer_id>"
    }
  ]
}