[Tracking] Support more OpenAI parameters for chat completion endpoint in REST API

[sudeepag]

Overview

The OpenAI chat completion API currently supports a number of parameters that are not currently supported in the mlc-llm REST API, such as temperature, stop sequences, etc. This issue tracks adding support for these parameters, so that they can be used by downstream dependencies of the REST API, such as Langchain integration.

Note that some parameters may require changes in the underlying chat module or llm_chat.

Action Items

• temperature
• top_p
• n
• stop
• max_tokens
• presence_penalty
• frequency_penalty

See the OpenAI API Reference for details on each of these parameters.

Links to Related Issues and PRs

• [Feature Request] Make v1/completions behave as OpenAI API #596
• [Feature Request] MLC Chat Rest API support stop params?  #674
• [API] Add GenerationConfig #1024
• [REST] OpenAI compatible Rest API #1107

[davidpissarra]

Concluded by #1107