[Model] Support Tele-FLM Model

[atone]

This PR adds support for Tele-FLM Model.

Tele-FLM (aka FLM-2) is a 52B open-sourced multilingual large language model that features a stable, efficient pre-training paradigm and enhanced factual judgement capabilities. Built upon the decoder-only transformer architecture, it has been trained on approximately 2T tokens. Tele-FLM demonstrates superior performances at its scale, and sometimes surpass larger models. In addition to sharing the model weights, we provide the core designs, engineering practices, and training details, anticipating their benefits for both academic and industrial communities.

The model collection can be found at Hugging Face (https://huggingface.co/collections/CofeAI/tele-flm-flm-2-669e4dbd2dbf53ccd2454304).

Veried that vllm serve CofeAI/FLM-2-52B-Instruct-2407 --trust-remote-code --chat-template /examples/tool_chat_template_teleflm.jinja -tp 2 works.

[github-actions]

👋 Hi! Thank you for contributing to the vLLM project.

💬 Join our developer Slack at https://slack.vllm.ai to discuss your PR in #pr-reviews, coordinate on features in #feat- channels, or join special interest groups in #sig- channels.

Just a reminder: PRs would not trigger full CI run by default. Instead, it would only run fastcheck CI which starts running only a small and essential subset of CI tests to quickly catch errors. You can run other CI tests on top of those by going to your fastcheck build on Buildkite UI (linked in the PR checks section) and unblock them. If you do not have permission to unblock, ping simon-mo or khluu to add you in our Buildkite org.

Once the PR is approved and ready to go, your PR reviewer(s) can run CI to test the changes comprehensively before merging.

To run CI, PR reviewers can either: Add ready label to the PR or enable auto-merge.

🚀

[Isotr0py]

Seems that here is the only difference with Llama, can you refactor the model implementation to reduce duplicate code just like glm.py and telechat2.py?

[Isotr0py]

Is really necessary to port this config? I think custom config from hf dynamic module can work in most of cases.

[jeejeelee]

FYI: The supported_lora_modules var has been deprecated, plz remove it

[jeejeelee]

Hi, remember to update the documentation, see: https://github.com/vllm-project/vllm/blob/main/docs/source/models/supported_models.md

[Isotr0py]

Don't forget updating tests/models/registry.py, otherwise test_initialization will fail! :)

[horizon94]

All the above suggestions have been adopted, and corresponding adjustments have been made:

1. The code now inherits from the llama class to minimize redundancy.
2. Updated tests/models/registry.py.
3. Updated the documentation.

The CI failure occurred in the Entrypoints Test, but this error appears unrelated to our code modifications.
Any suggestion? :)

[DarkLight1337]

Can you merge from main to fix the CI failure?

[DarkLight1337]

CI failures look unrelated, merging. Thanks for your effort!