[Speculative Decoding]【Hackathon 10th Spring No.49】Adapt ngram_match and hybrid_mtp_ngram gpu kernels

[NKNaN]

Motivation

rfc: PaddlePaddle/community#1213

Modifications

• 实现方式：两个 kernel。
  • 第一阶段：count_and_find_candidate_kernel，网格为 <<<max_batch_size+1, 1024>>>。
    • block 0 用 BlockReduce 统计全局 unprocessed_batch_size。
    • block 1..N 各自负责一个 batch 并行执行候选查找（input_ids / pre_ids）。
  • 第二阶段：truncate_candidate，<<<1, 1024>>>，统一按 threshold 做截断和写回。
    • 该阶段使用 CUB BlockScan 做前缀和（processed_batch_size / sum_token_num），用于计算每个 batch 的可分配 token 上限并完成截断。

Usage or Command

Accuracy Tests

https://github.com/NKNaN/FastDeploy_ngram_match_kernel

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!