add document for deepseek large EP

[hust17yixuan]

What this PR does / why we need it?

This PR presents a large-EP deployment solution based on vllm-ascend, using DeepSeek as an example. It outlines the end-to-end workflow for model deployment and serves as a reference for developers.

Does this PR introduce any user-facing change?

No

How was this patch tested?

[gemini-code-assist]

Code Review

This pull request adds documentation for deploying a large Expert Parallelism (EP) model using a distributed DP server. The documentation is comprehensive, but I've found a few issues in the code examples that could cause problems for users.
Specifically, a Python script example has incorrect data types for variables, which will lead to runtime errors. There's also a likely copy-paste error in a shell command example, and another command example is illustrative rather than a concrete, runnable command, which might be confusing. Finally, there's a syntax error in a markdown note block that will likely break rendering.
I've provided suggestions to fix these issues to improve the clarity and correctness of the documentation.

[gemini-code-assist]

The placeholder values for dp_size, dp_size_local, dp_rank_start, dp_port, and engine_port are strings, but they are used in arithmetic operations or as port numbers, which requires integers. This will cause a TypeError when the script is run. The placeholder values should be integers, and it would be helpful to provide example integer values.

Suggested change

	dp_size = "all dp size for decode/prefill"
	dp_size_local = "dp size for current node"
	dp_rank_start = "dp beginning number for current node"
	dp_ip = "master node ip"
	dp_port = "port used to communication"
	engine_port = "the beginning port for all dp group in current node"
	dp_size = 8 # Example: total number of DP workers for decode/prefill
	dp_size_local = 2 # Example: number of DP workers on the current node
	dp_rank_start = 0 # Example: starting DP rank for the current node
	dp_ip = "192.168.1.100" # Example: master node IP
	dp_port = 12345 # Example: port used for communication
	engine_port = 8000 # Example: the starting port for all DP groups on the current node

[hust17yixuan]

fixed

[gemini-code-assist]

There appears to be a copy-paste error in the example command for gen_ranktable.sh. The IP for decoder_node1_local_ip is repeated, which is likely incorrect and should probably be for two different decoder nodes (e.g., decoder_node1_local_ip and decoder_node2_local_ip). This could lead to misconfiguration for users following this guide.

Suggested change

	bash gen_ranktable.sh --ips prefiller_node1_local_ip prefiller_node2_local_ip decoder_node1_local_ip decoder_node1_local_ip \
	bash gen_ranktable.sh --ips prefiller_node1_local_ip prefiller_node2_local_ip decoder_node1_local_ip decoder_node2_local_ip \

[hust17yixuan]

fixed

[gemini-code-assist]

The example for running load_balance_proxy_server_example.py is illustrative and contains placeholders that are not valid command-line arguments (e.g., prefiller node1 local ip, engine_port ...). This could be confusing for users. It would be more helpful to provide a complete, concrete example that they can adapt.

Suggested change

	python load_balance_proxy_server_example.py \
	--port "proxy port" \
	--host 0.0.0.0 \
	--prefiller-hosts \
	prefiller node1 local ip \
	prefiller node2 local ip \
	--prefiller-ports \
	engine_port engine_port \
	--decoder-hosts \
	decoder node1 local ip \
	decoder node1 local ip \
	decoder node2 local ip \
	decoder node2 local ip \
	--decoder-ports \
	engine_port ... engine_port + dp_size_local \
	engine_port ... engine_port + dp_size_local \
	python load_balance_proxy_server_example.py \
	--port 8000 \
	--host 0.0.0.0 \
	--prefiller-hosts \
	192.168.1.101 \
	192.168.1.102 \
	--prefiller-ports \
	9000 9001 \
	--decoder-hosts \
	192.168.1.103 \
	192.168.1.103 \
	192.168.1.104 \
	192.168.1.104 \
	--decoder-ports \
	9002 9003 9004 9005

[hust17yixuan]

fixed

[gemini-code-assist]

The syntax for the note block is incorrect. It uses ::::{note} and :::: which will likely cause rendering issues. The correct MyST syntax for a note directive is :::{note} and :::.

Suggested change

	:::{note}
	Each node local ip should repeat the same times as its '**dp_size_local**', at the same time, each node has the same number of ports as '**dp_size_local**', and ther ports increase sequentially starting from '**engine_port**'.
	:::
	:::{note}
	Each node local ip should repeat the same times as its '**dp_size_local**', at the same time, each node has the same number of ports as '**dp_size_local**', and ther ports increase sequentially starting from '**engine_port**'.
	:::