[Model] Add support for DeepSeek-V2 Model

[rickzx]

This PR implements the DeepSeek-V2 Model architecture: https://huggingface.co/deepseek-ai/DeepSeek-V2-Lite/blob/main/modeling_deepseek.py.

The notable changes from the common LLM architecture includes:

• Multihead Latent Attention (MLA)
• Yarn Rotary Positional Embeddings
• DeepSeekMoE

Example execution on M2 ultra:

% mlc_llm chat ../models/DeepSeek-V2-Lite-Chat-MLC-q0f16 --model-lib ../models/DeepSeek-V2-Lite-Chat-MLC-q
0f16/model.dylib
>>> who are you?
 I am an AI assistant created by DeepSeek to be helpful and harmless.

TODO:

• Currently the model architecture only supports Deepseek-V2-Lite. To support Deepseek-V2, we also need to support the group_limited_greedy strategy.
• Support tensor parallel > 1.

[fengyang95]

Hello @rickzx , I noticed that the PR has been merged, and I wanted to ask if the todo items mentioned have been completed. Is support now available for deepseek-v2 (not the lite version)?

[rickzx]

Hello @rickzx , I noticed that the PR has been merged, and I wanted to ask if the todo items mentioned have been completed. Is support now available for deepseek-v2 (not the lite version)?

Hi @fengyang95, the TODO items mentioned above hasn't been completed yet. I plan to finish those items asap, hopefully by end of week. The non-lite version is not supported with the current version