* Fixes #26: A Model is now retrieved by the backends.get_model_for(model_spec) method which performs a unification operation with existing ModelSpecs from a model registry. The first unifying model spec is returned or the one given retained. A ModelSpec must define a "backend" (name) to properly load a Model. The backend name must match a backend file like <name>_api.py.

Changes:
- Backends now provide Models
- Models are defined by ModelSpecs
- ModelSpecs are described in a model_registry.json
- generation arguments (temp, max_tokens) are directly attached to Model
- backends are now lazy loaded and the one to be used must be specified in the ModelSpec

New Feature:
- now the benchmark will try to parse the -m option as json to create a ModelSpec (names also still work)
- this looks like: python3 scripts/cli.py run -g taboo -m "{'model_name':'gpt3-turbo','backend':'openai'}"
- note that single-quotes must be specified (these will be replaced to create proper json)

Aside:
- adjusted all pre-defined bechmark games to newly introduced classes
- remove text-davinci-003 (not listed anymore in https://api.openai.com/v1/models)
- prototyping: additional model specs can be defined in a model_registry_custom.json (not version controlled)
- prototyping: use model.set_gen_args(arg0=1,arg1=2) to set generation arguments

* Extract `GameScorer` from the `GameMaster` scoring code
* GameBenchmarks must now implement `create_game_scorer` to return a `GameScorer`

* issue #47: add option to set max_token to be generated

* issue #47: rename max_token option to -l (limit, length)

- CodeLlama-70b-Instruct-hf
- openchat-3.5-0106
- openchat-3.5-1210 
- Nous-Hermes-2-Mixtral-8x7B-DPO

* HF backend refactor (#37)

- Model settings are now stored in external JSON model registry
- Two new prototyping methods to check messages for issues and context limit
- General Exception to handle exceeding context
- Gated model (Llama2) handling integrated into single HF backend
- New and updated documentation
- use huggingfaces' cache lookup mechanism

Co-authored-by: Jonathan Jordan <knawurzelkopp@hotmail.de>

* add option to specify filename for instances in GameInstanceGenerator.generate(), default is set to instances.json to allow for backward compatibility

* inline no where else used storing function

---------

Co-authored-by: Anne Beyer <anne.beyer@uni-potsdam.de>

* Fixing typos in framework
* Added a reprompting mechanism to the dialogue game master

Fixes #34

- now clem automatically assumes self-play when only a single model is given (and the game is multi-player)
- removed the option to run all games (as this simplifies the code and is now done in a pipeline script)
- improved cli usage `python3 scripts/cli.py run -g <game> -m <model1> [<model2>]`
- pipeline scripts and README have been updated (no changes in the games)

* [clemgame] store results grouped by model pairing (instead of game): results->pair->game
* [eval] swap order of game and model in results structure

---------

Co-authored-by: briemadu <madureiralasota@uni-potsdam.de>

* Add openchat_3.5, Yi-34B-Chat, Orca-2-13b model support; Add custom openchat and ChatML chat templates; Add 'slow' tokenizer usage handling

* Add culling of openchat EOS token at end of outputs

* Remove Yi-34B-Chat and Orca-2-13b from supported models list due to issues found in testing

* Add Yi-34B-Chat output processing; Add Yi-34B-Chat to supported models list

* Add DeepSeek 7B/67B, Tulu 7B/70B and Mixtral 8x7B models

* Changed transformers version requirement to 4.36.0

* Changed accelerate version requirement to 0.25.0

* Add output EOS culling for DeepSeek

* Add culling of empty iniitla system message for Mistral models compatibility

* Add SUSTech/SUS-Chat-34B to huggingface_local_api.py

* Add SUSTech/SUS-Chat-34B output EOS culling

* Started model checking cli script

* Finish initial_hf_check.py; Add HowTo doc on adding models to the HF backend

* Add openchat_3.5, Yi-34B-Chat, Orca-2-13b model support; Add custom openchat and ChatML chat templates; Add 'slow' tokenizer usage handling

* Add culling of openchat EOS token at end of outputs

* Remove Yi-34B-Chat and Orca-2-13b from supported models list due to issues found in testing

* Add Yi-34B-Chat output processing; Add Yi-34B-Chat to supported models list

* Add DeepSeek 7B/67B, Tulu 7B/70B and Mixtral 8x7B models

* Changed transformers version requirement to 4.36.0

* Changed accelerate version requirement to 0.25.0

* Add output EOS culling for DeepSeek