| transformer类onnx算子插件支持
Co-authored-by: xuyang12138<xuyang270@huawei.com>
# message auto-generated for no-merge-commit merge:
!539 merge master into master
transformer类onnx算子插件支持
Created-by: yanke-xu
Commit-by: xuyang12138
Merged-by: cann-robot
Description: ## 描述
transformer类onnx算子插件支持,包含以下onnx算子类型:
1. NPUFlashAttention
2. NPUIncreFlashAttention
3. NPUPromptFlashAttention
4. EmbeddingBag
5. FillWindowCache
6. NPUMultiHeadAttention
7. NPUFusedAttentionScoreFwd
8. NPUFusedAttentionScore
9. NPUMaskedSoftmaxWithRelPosBias
10. NPUScaledMaskedSoftmax
11. TfIdfVectorizer
12. NPUMoeComputeExpertTokens
13. NPUMoeFinalizeRouting
14. NPUMoeFinalizeRoutingV2
15. NPUMoeGatingTopKSoftmax
16. NPUMoeInitRouting
17. NPURotaryPositionEmbedding
## 关联的Issue
None
## 测试
流水线验证
## 文档更新
None
## 类型标签
<!-- [x] 表示选中 -->
- [ ] Bug修复
- [x] 新特性
- [ ] 性能优化
- [ ] 文档更新
- [ ] 其他,请描述:
See merge request: cann/ops-transformer!539 | 4 个月前 |