saymrwulf/onnxruntime
1
0
Fork 0
mirror of https://github.com/saymrwulf/onnxruntime.git synced 2026-07-27 20:02:15 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

Remove file-scope non-constant static variables to support multiple inference sessions (#10481)

Browse source 
* Changed file-scope static variables to automatic variables or function-scope static const.

* Reduce load time overhead by using constexpr.

* Use node indices instead of node names to track inserted, deleted and changed nodes.

Co-authored-by: Satya Jandhyala <sajandhy@microsoft.com>
This commit is contained in:
satyajandhyala 2022-02-10 13:31:12 -08:00 committed by GitHub
parent 4d6d4dfb9d
commit eba730500f
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
2 changed files with 552 additions and 516 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

1059
onnxruntime/core/optimizer/propagate_cast_ops.cc
View file

File diff suppressed because it is too large Load diff

9
onnxruntime/core/optimizer/propagate_cast_ops.h
View file

@ -15,6 +15,14 @@ Propagate FP16 Cast operations up the graph and FP32 Cast operations down the gr
*/
class PropagateCastOps : public GraphTransformer {
public:
/*
* The collection FP16AllowOps, specifies for a given propagate_cast_ops level, a collection of node op_types that
* the code is allowed to propage Cast operations across. The user may specify a custom list of optypes using level 0.
* The opcodes are split into multiple levels. Cast propagation is done based on the level. Level 2 op code
* list includes Level 1 list also.
*/
typedef std::vector<std::unordered_set<std::string>> FP16AllowOps;
PropagateCastOps(GraphTransformerConfiguration::PropagateCastOpsConfiguration::Strategy strategy =
GraphTransformerConfiguration::PropagateCastOpsConfiguration::Strategy::FloodFill,
size_t level = 0, const std::vector<std::string>& allow_list = {},
@ -24,6 +32,7 @@ class PropagateCastOps : public GraphTransformer {
private:
size_t level_;
FP16AllowOps fp16_allow_ops_;
GraphTransformerConfiguration::PropagateCastOpsConfiguration::Strategy strategy_;
};