Support stash_type attribute for onnx.LayerNormalization

llvm · Nov 24, 2024 · 367e6d1 · 367e6d1
1 parent 878f992
commit 367e6d1
Show file tree

Hide file tree

Showing 2 changed files with 72 additions and 23 deletions.
diff --git a/lib/Conversion/TorchOnnxToTorch/DefaultDomainGtoP.cpp b/lib/Conversion/TorchOnnxToTorch/DefaultDomainGtoP.cpp
@@ -2543,7 +2543,34 @@ void mlir::torch::onnx_c::populateDefaultDomainGtoP(
             binder.s64IntegerAttr(stashType, "stash_type", 1))
           return failure();
 
+        std::optional<int64_t> stashTypeIntTorch =
+            onnxDtypeIntToTorchDtypeInt(stashType);
+        if (!stashTypeIntTorch.has_value())
+          return rewriter.notifyMatchFailure(
+              binder.op, "unimplemented support for the given stash_type");
+        FailureOr<Type> stashDtype = Torch::getTypeForScalarType(
+            binder.op->getContext(),
+            (torch_upstream::ScalarType)stashTypeIntTorch.value());
+        if (failed(stashDtype))
+          return failure();
+
+        // Convert dtype if stash_type is different from input dtype
         auto xType = cast<Torch::ValueTensorType>(x.getType());
+        Value cstFalse =
+            rewriter.create<Torch::ConstantBoolOp>(binder.getLoc(), false);
+        Value none = rewriter.create<Torch::ConstantNoneOp>(binder.getLoc());
+        if (*stashDtype != xType.getOptionalDtype()) {
+          auto newXType =
+              xType.getWithSizesAndDtype(xType.getOptionalSizes(), *stashDtype);
+          Value dtypeValue = rewriter.create<Torch::ConstantIntOp>(
+              binder.getLoc(),
+              rewriter.getI64IntegerAttr(stashTypeIntTorch.value()));
+          x = rewriter.create<Torch::AtenToDtypeOp>(
+              binder.getLoc(), newXType, x, /*dtype=*/dtypeValue,
+              /*non_blocking=*/cstFalse, /*copy=*/cstFalse,
+              /*memory_format=*/none);
+        }
+
         Value constEpsilon = rewriter.create<Torch::ConstantFloatOp>(
             binder.getLoc(), rewriter.getType<Torch::FloatType>(),
             rewriter.getF64FloatAttr(epsilon));
@@ -2566,33 +2593,43 @@ void mlir::torch::onnx_c::populateDefaultDomainGtoP(
             Torch::ListType::get(Torch::IntType::get(binder.op->getContext())),
             normalized);
 
+        SmallVector<int64_t> reducedShape(rank, 1);
+        for (int64_t i = 0; i < axis; i++)
+          reducedShape[i] = xShape[i];
+        auto reducedType =
+            xType.getWithSizesAndDtype(reducedShape, *stashDtype);
+        auto y = rewriter.create<Torch::AtenNativeLayerNormOp>(
+            binder.getLoc(), yType, /*meanType=*/reducedType,
+            /*invStdDevType=*/reducedType, x, normalized_shape,
+            scale, b, constEpsilon);
+
         int64_t numResults = binder.op->getNumResults();
         if (numResults == 1) {
-          SmallVector<int64_t> reducedShape(rank, 1);
-          for (int64_t i = 0; i < axis; i++)
-            reducedShape[i] = xShape[i];
-          auto reducedType = xType.getWithSizesAndDtype(
-              reducedShape, xType.getOptionalDtype());
-          Value y = rewriter
-                        .create<Torch::AtenNativeLayerNormOp>(
-                            binder.getLoc(), yType, /*meanType=*/reducedType,
-                            /*invStdDevType=*/reducedType, x, normalized_shape,
-                            scale, b, constEpsilon)
-                        .getResult0();
-          rewriter.replaceOp(binder.op, y);
+          rewriter.replaceOp(binder.op, y.getResult0());
           return success();
         }
-        if (numResults == 3) {
-          if (binder.tensorResultTypeAtIndex(meanType, 1) ||
-              binder.tensorResultTypeAtIndex(invStdDevType, 2))
-            return failure();
-          rewriter.replaceOpWithNewOp<Torch::AtenNativeLayerNormOp>(
-              binder.op, yType, meanType, invStdDevType, x, normalized_shape,
-              scale, b, constEpsilon);
-          return success();
+
+        Value meanOutput = y.getResult1();
+        Value varOutput = y.getResult2();
+        // Convert meanType and varType back if stash_dtype is different
+        if (binder.tensorResultTypeAtIndex(meanType, 1) ||
+            binder.tensorResultTypeAtIndex(invStdDevType, 2))
+          return failure();
+        if (*stashDtype != meanType.getOptionalDtype()) {
+          Value constDtype = Torch::getDtypeIntValueForType(
+              rewriter, binder.getLoc(), meanType.getDtype());
+          meanOutput = rewriter.create<Torch::AtenToDtypeOp>(
+              binder.getLoc(), meanType, meanOutput, /*dtype=*/constDtype,
+              /*non_blocking=*/cstFalse, /*copy=*/cstFalse,
+              /*memory_format=*/none);
+          varOutput = rewriter.create<Torch::AtenToDtypeOp>(
+              binder.getLoc(), invStdDevType, varOutput, /*dtype=*/constDtype,
+              /*non_blocking=*/cstFalse, /*copy=*/cstFalse,
+              /*memory_format=*/none);
         }
-        return rewriter.notifyMatchFailure(
-            binder.op, "Unimplemented: expected either 1 or 3 results");
+        rewriter.replaceOp(binder.op, {y.getResult0(), meanOutput, varOutput});
+
+        return success();
       });
   patterns.onOp("LeakyRelu", 1,
                 [](OpBinder binder, ConversionPatternRewriter &rewriter) {

diff --git a/lib/Dialect/Torch/Transforms/DecomposeComplexOps.cpp b/lib/Dialect/Torch/Transforms/DecomposeComplexOps.cpp
@@ -6635,7 +6635,7 @@ class DecomposeAtenNativeLayerNormOp
     Location loc = op.getLoc();
     auto context = op.getContext();
 
-    auto inputTy = cast<BaseTensorType>(op.getInput().getType());
+    auto inputTy = cast<ValueTensorType>(op.getInput().getType());
     if (!inputTy.hasSizes())
       return rewriter.notifyMatchFailure(
           op, "input tensor should have known sizes.");
@@ -6690,6 +6690,18 @@ class DecomposeAtenNativeLayerNormOp
         loc, inputTy, inputRsqrtVar, op.getInput());
     Value inputNormalized = rewriter.create<AtenMulTensorOp>(
         loc, inputTy, inputZeroMean, inputRsqrtVarExpanded);
+    // Convert resultType if dtype is different
+    auto resultTensorType =
+        dyn_cast<ValueTensorType>(op.getResult(0).getType());
+    if (inputTy.getDtype() != resultTensorType.getDtype()) {
+      Value dtypeValue = Torch::getDtypeIntValueForType(
+          rewriter, loc, resultTensorType.getDtype());
+      Value cstFalse = rewriter.create<Torch::ConstantBoolOp>(loc, false);
+      inputNormalized = rewriter.create<Torch::AtenToDtypeOp>(
+          loc, resultTensorType, inputNormalized,
+          /*dtype=*/dtypeValue, /*non_blocking=*/cstFalse, /*copy=*/cstFalse,
+          /*memory_format=*/none);
+    }
     Value out = rewriter.create<TensorStaticInfoCastOp>(
         loc, op.getResult(0).getType(), inputNormalized);