LocalAI Config files

trafficstars

This PR adds a first step into having a configuration mechanism for LocalAI.

Fixes: https://github.com/go-skynet/LocalAI/issues/28

It introduces two way of configuring LocalAI "virtual models": a single YAML file containing all the user-defined models and their mapping, that can be specified via CLI, or by creating multiple YAML files inside the models folder.

A config file might be:

- name: gpt-3.5-turbo
  parameters:
    model: ggml-gpt4all-j
  context_size: 1024
  threads: 10
  stopwords:
  - "HUMAN:"
  - "GPT:"
  roles:
    user: "HUMAN:"
    system: "GPT:"
  template:
    completion: completion
    chat: ggml-gpt4all-j
- name: list2
  parameters:
    model: testmodel
  context_size: 512
  threads: 10
  stopwords:
  - "HUMAN:"
  - "### Response:"
  roles:
    user: "HUMAN:"
    system: "GPT:"
  template:
    completion: completion
    chat: ggml-gpt4all-j

A model declaration, that can be inside the models directory can be:

name: gpt4all
parameters:
  model: testmodel # A path to a model file inside the models folder
context_size: 512
threads: 10
stopwords:
- "HUMAN:"
- "### Response:"
cutstrings:
- ...
trimspace:
- ...
roles:
  user: "HUMAN:"
  system: "GPT:"
template:
  completion: completion
  chat: ggml-gpt4all-j

Every model declared will be available for inference and returned to the models list.

Fields:

Inside parameters can be specified all the OpenAI request fields. That includes, temperature, top_p, top_k
Inside template can be specified files to be used for endpoints to templates requests with. Files must end in "tmpl"

Supersedes https://github.com/go-skynet/LocalAI/pull/63