add DeepSeek-R1 tutorial.

[Gongdayao]

What this PR does / why we need it?

This PR adds tutorials for the DeepSeeK-R1 series models, including the A2 and A3 series, and provides accuracy validation results.

Does this PR introduce any user-facing change?

How was this patch tested?

• vLLM version: v0.12.0
• vLLM main: https://github.com/vllm-project/vllm/commit/v0.12.0

[gemini-code-assist]

Code Review

This pull request adds a new tutorial for deploying the DeepSeek-R1 model. The documentation is well-structured and provides detailed instructions for both single-node and multi-node deployments. However, I found some critical inconsistencies in the model name within the provided commands, which will cause them to fail. Correcting these will significantly improve the user experience of following this tutorial.

[gemini-code-assist]

The model path path/DeepSeek-R1-w8a8 uses w8a8 in lowercase, which is inconsistent with the model name DeepSeek-R1-W8A8 used in the download link and the vllm serve commands. This will cause a 'file not found' error on case-sensitive filesystems.

Please correct this to DeepSeek-R1-W8A8 to ensure the command works as expected. Similar inconsistencies are present on lines 6, 18, and 291 and should also be corrected.

Suggested change

	lm_eval \
	--model local-completions \
	--model_args model=path/DeepSeek-R1-w8a8,base_url=http://<node0_ip>:<port>/v1/completions,tokenized_requests=False,trust_remote_code=True \
	--tasks gsm8k \
	--output_path ./
	lm_eval \
	--model local-completions \
	--model_args model=path/DeepSeek-R1-W8A8,base_url=http://<node0_ip>:<port>/v1/completions,tokenized_requests=False,trust_remote_code=True \
	--tasks gsm8k \
	--output_path ./

[gemini-code-assist]

The model path path/DeepSeek-R1-w8a8 is inconsistent with the official model name DeepSeek-R1-W8A8 from the download link. This will cause a 'file not found' error. Please correct it to DeepSeek-R1-W8A8.

Suggested change

	vllm bench serve --model path/DeepSeek-R1-w8a8 --dataset-name random --random-input 200 --num-prompt 200 --request-rate 1 --save-result --result-dir ./
	vllm bench serve --model path/DeepSeek-R1-W8A8 --dataset-name random --random-input 200 --num-prompt 200 --request-rate 1 --save-result --result-dir ./

[github-actions]

👋 Hi! Thank you for contributing to the vLLM Ascend project. The following points will speed up your PR merge:‌‌

• A PR should do only one thing, smaller PRs enable faster reviews.
• Every PR should include unit tests and end-to-end tests ‌to ensure it works and is not broken by other future PRs.
• Write the commit message by fulfilling the PR description to help reviewer and future developers understand.

If CI fails, you can run linting and testing checks locally according Contributing and Testing.

[1092626063]

provide docker run command directly, like "#4399"

[1092626063]

provide docker run command by refering A3 series

[1092626063]

/lgtm