[Cherry-Pick] [FDConfig] Unify num_experts_per_tok to moe_k in ModelConfig for MoE model compatibility(#7509)

[xyxinyang]

Motivation

💡 If this PR is a Cherry Pick, the PR title needs to follow the format by adding the [Cherry-Pick] label at the very beginning and appending the original PR ID at the end. For example, [Cherry-Pick][CI] Add check trigger and logic(#5191)

💡 如若此PR是Cherry Pick，PR标题需遵循格式，在最开始加上[Cherry-Pick]标签，以及最后面加上原PR ID，例如[Cherry-Pick][CI] Add check trigger and logic(#5191)

Unify num_experts_per_tok to moe_k in ModelConfig for MoE model compatibility. This enables R3 support for models like Qwen3VLMOE, DeepSeek V3, and other MoE models that use num_experts_per_tok instead of moe_k in their config.

Modifications

• Added unified field mapping in ModelConfig (fastdeploy/config.py)

Usage or Command

N/A

Accuracy Tests

N/A

Checklist

• Add at least a tag in the PR title.
  • Tag list: [[FDConfig],[APIServer],[Engine], [Scheduler], [PD Disaggregation], [Executor], [Graph Optimization], [Speculative Decoding], [RL], [Models], [Quantization], [Loader], [OP], [KVCache], [DataProcessor], [BugFix], [Docs], [CI], [Optimization], [Feature], [Benchmark], [Others], [XPU], [HPU], [GCU], [DCU], [Iluvatar], [Metax]]
  • You can add new tags based on the PR content, but the semantics must be clear.
• Format your code, run pre-commit before commit.
• Add unit tests. Please write the reason in this PR if no unit tests.
• Provide accuracy results.
• If the current PR is submitting to the release branch, make sure the PR has been submitted to the develop branch, then cherry-pick it to the release branch with the [Cherry-Pick] PR tag.

[paddle-bot]

Thanks for your contribution!

[PaddlePaddle-bot]

🤖 AI Code Review | 2026-04-20 16:03 CST

📋 Review 摘要

PR 概述：将 MoE 模型 config 中的 num_experts_per_tok 统一映射为 moe_k，以支持 Qwen3VLMOE、DeepSeek V3 等模型的兼容性。
变更范围：fastdeploy/config.py — ModelConfig.override_name_from_config 方法
影响面 Tag：FDConfig

问题

级别	文件	概述
🟡 建议	多个模型文件	下游消费方仍直接使用 num_experts_per_tok，建议统一迁移为 moe_k

总体评价

变更逻辑正确，与同方法内 moe_num_experts/moe_num_shared_experts 的映射模式一致。not hasattr(self, "moe_k") 的守卫条件也恰当（moe_k 不在 PRETRAINED_INIT_CONFIGURATION 中，因此用 hasattr 比 is None 更合理）。建议后续跟进将下游直接引用 num_experts_per_tok 的代码统一迁移到 moe_k，以保持代码风格一致性。

[PaddlePaddle-bot]

🟡 建议 映射逻辑本身正确，但目前有多个下游文件仍直接使用 fd_config.model_config.num_experts_per_tok 而非统一后的 moe_k：

• fastdeploy/model_executor/models/deepseek_v3.py:173
• fastdeploy/model_executor/models/qwen3moe.py:68
• fastdeploy/model_executor/models/gpt_oss.py:119
• fastdeploy/model_executor/models/glm4_moe.py:178
• fastdeploy/model_executor/layers/moe/routing_indices_cache.py:170（对 Glm4Moe 的特殊分支）

由于映射代码只赋值 self.moe_k = self.num_experts_per_tok 而未删除原属性，这些引用不会报错，不阻塞合入。但建议在后续 PR 中将上述引用统一迁移为 moe_k，与 ernie4_5_moe.py 等已使用 moe_k 的模型保持一致，减少维护歧义。