[Misc]: 特性梳理

[wangshuai09]

特性	分析	适配情况
PageAttention	decoding 阶段 attention
FlashAttention	prefilling 阶段 attention
TP	tensor多卡并行
Multi-Host tensor-parallel	多节点tp并行，需适配vllm ray_executor
FlashDecoding	attention算子部分场景优化,torch_npu中page attention接口已支持，未适配，flash attention接口未支持
MoE	混合专家模型，需适配两个算子
Compilation cache	编译缓存，依赖torch.compiler
Prefix caching	需slot_indices部分适配
Vision language model	需逐模型适配，部分支持
Speculative decoding	投机掩码
Chunked prefills	需slot_indices部分调试
Lora	需实现6个triton算子
MultiLora	需实现6个triton算子
Int8/AWQ/GPTQ量化	需实现量化相关算子
BeamSearch	优化的搜索方法
Custom_op融合算子	需实现acend算子或调用torch_npu相关算子
310p适配	以上已支持特性均为910b环境测试并适配，310p适配需调试torch_npu attention接口，后续特性支持需同步适配310p

Before submitting a new issue...

• Make sure you already searched for relevant issues, and asked the chatbot living at the bottom right corner of the documentation page, which can answer lots of frequently asked questions.

[Shenggao190]

你好，我想问一下，量化相关部分还需要多久可以适配？

[github-actions]

This issue has been automatically marked as stale because it has not had any activity within 90 days. It will be automatically closed if no further activity occurs within 30 days. Leave a comment if you feel this issue should remain open. Thank you!