Add Anyscale Endpoint support and Llama Tokenizer

[kylehh]

Problem

1. No support for open source models
2. No support for open source tokenizers

Solution

1. Add Anyscale Endpoint LLM client which hosts Llama2 , Mistrial and Zephyr models.
2. Add Llama Tokenizer based on huggingface's tokenizers library

Type of Change

• Bug fix (non-breaking change which fixes an issue)
• [x ] New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• This change requires a documentation update
• Infrastructure change (CI configs, etc)
• Non-code change (docs, etc)
• None of the above: (explain here)

Test Plan

Describe specific steps for validating this change.

1. Get your HF and Anyscale tokens and add them in the config/anyscale.yaml
2. canopy start --config anyscale.yaml

[igiloh-pinecone]

@kylehh Thank you very much for your contribution!! This is highly appreciated!

Please see a few comments. The main 3 issues:

1. We plan to merge Upgrade openai sdk to v.1.2.3 #171 today, which adds a base_url init argument to OpenAILLM class. With that change, the implementation of AnyscaleLLM could be as easy as inheriting OpenAILLM and passing hard coded base_url to super().__init__().
2. The new classes, AnyscaleLLM and LlamaTokenizer require their own unit classes before this PR can be merged.
3. Let's figure out together if it's really required for every Canopy user to generate a HF identification token just for the Tokenizer, or can we find a more user friendly solution.

[igiloh-pinecone]

HI @kylehh,

As we discussed over slack - there are some changes needed for the Tokenizer, please see suggested changes in kylehh#1.

In addition, there a few more issues to solve, namely the unit test for AnyscaleLLM and the linter tests

[igiloh-pinecone]

LGTM!

Thank you @kylehh for all of your efforts. Merging.