[doc] fix qwen35 gkd deploy

[hjh0119]

No description provided.

[gemini-code-assist]

Code Review

This pull request updates the documentation for Qwen3.5 best practices (both Chinese and English versions) to use swift deploy instead of vllm serve for launching the teacher server during GKD training. The review feedback points out that the --max_logprobs parameter should be prefixed with --vllm_ (i.e., --vllm_max_logprobs) to be correctly parsed by swift deploy. Additionally, it recommends correcting the definition of GKD to "Generalized Knowledge Distillation" in the English documentation.

Important

The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.

[gemini-code-assist]

In swift deploy, vLLM-specific arguments should be prefixed with --vllm_ to be correctly parsed and passed to the vLLM engine. Please use --vllm_max_logprobs instead of --max_logprobs.

Suggested change

	--max_logprobs 64
	--vllm_max_logprobs 64

[gemini-code-assist]

GKD stands for Generalized Knowledge Distillation (from the GKD paper) rather than "General Knowledge Distillation". Let's correct this terminology for accuracy.

Suggested change

	LoRA training with GKD (General Knowledge Distillation), using Qwen3.5-9B as the teacher model. First, launch the teacher server with `swift deploy` (alternatively, use the `--teacher_model` parameter to load the model directly):
	LoRA training with GKD (Generalized Knowledge Distillation), using Qwen3.5-9B as the teacher model. First, launch the teacher server with swift deploy (alternatively, use the --teacher_model parameter to load the model directly):

[gemini-code-assist]

In swift deploy, vLLM-specific arguments should be prefixed with --vllm_ to be correctly parsed and passed to the vLLM engine. Please use --vllm_max_logprobs instead of --max_logprobs.

Suggested change

	--max_logprobs 64
	--vllm_max_logprobs 64