[MLInlineAdvisor] Changing creation of TensorSpec in a type agnostic manner.

[svkeerthy]

This change would enable using ir2vec::Embedding which are float vectors in ML Inliner.

[llvmbot]

@llvm/pr-subscribers-llvm-analysis

Author: S. VenkataKeerthy (svkeerthy)

Changes

This change would enable using ir2vec::Embedding which are float vectors in ML Inliner.

---

Full diff: https://github.com/llvm/llvm-project/pull/141161.diff

1 Files Affected:

• (modified) llvm/lib/Analysis/DevelopmentModeInlineAdvisor.cpp (+2-2)

diff --git a/llvm/lib/Analysis/DevelopmentModeInlineAdvisor.cpp b/llvm/lib/Analysis/DevelopmentModeInlineAdvisor.cpp
index 2feedd2ff40af..e7e8f2ac1ff25 100644
--- a/llvm/lib/Analysis/DevelopmentModeInlineAdvisor.cpp
+++ b/llvm/lib/Analysis/DevelopmentModeInlineAdvisor.cpp
@@ -261,8 +261,8 @@ static const std::vector<TensorSpec> TrainingOnlyFeatures{
 static const std::vector<TensorSpec> getInputFeatures() {
   std::vector<TensorSpec> InputSpecs;
   for (size_t I = 0; I < NumberOfFeatures; ++I)
-    InputSpecs.push_back(TensorSpec::createSpec<int64_t>(
-        TFFeedPrefix + FeatureMap[I].name(), FeatureMap[I].shape()));
+    InputSpecs.push_back(
+        TensorSpec(TFFeedPrefix + FeatureMap[I].name(), FeatureMap[I]));
   append_range(InputSpecs, TrainingOnlyFeatures);
   return InputSpecs;
 }

[mtrofin]

lgtm, can you change the subject to say this is in the inline advisor, otherwise it reads like all TensorSpecs do that.

[mtrofin]

Failures seem unrelated, merging.