diff --git a/mlir/lib/Conversion/TosaToLinalg/TosaToLinalg.cpp b/mlir/lib/Conversion/TosaToLinalg/TosaToLinalg.cpp
--- a/mlir/lib/Conversion/TosaToLinalg/TosaToLinalg.cpp
+++ b/mlir/lib/Conversion/TosaToLinalg/TosaToLinalg.cpp
@@ -339,6 +339,21 @@
   }
 };
 
+// At the codegen level any identity operations should be removed. Any cases
+// where identity is load-bearing (e.g. cross device computation) should be
+// handled before lowering to codegen.
+template <typename SrcOp>
+class IdentityNConverter : public OpRewritePattern<SrcOp> {
+public:
+  using OpRewritePattern<SrcOp>::OpRewritePattern;
+
+  LogicalResult matchAndRewrite(SrcOp op,
+                                PatternRewriter &rewriter) const final {
+    rewriter.replaceOp(op, op.getOperation()->getOperands());
+    return success();
+  }
+};
+
 } // namespace
 
 void mlir::tosa::populateTosaToLinalgOnTensorsConversionPatterns(
@@ -358,6 +373,7 @@
       PointwiseConverter<tosa::GreaterEqualOp>,
       PointwiseConverter<tosa::MaximumOp>, PointwiseConverter<tosa::MinimumOp>,
       PointwiseConverter<tosa::CeilOp>, PointwiseConverter<tosa::FloorOp>,
-      PointwiseConverter<tosa::ClampOp>, PointwiseConverter<tosa::ReluNOp>>(
-      context);
+      PointwiseConverter<tosa::ClampOp>, PointwiseConverter<tosa::ReluNOp>,
+      IdentityNConverter<tosa::IdentityOp>,
+      IdentityNConverter<tosa::IdentityNOp>>(context);
 }
diff --git a/mlir/test/Conversion/TosaToLinalg/tosa-to-linalg.mlir b/mlir/test/Conversion/TosaToLinalg/tosa-to-linalg.mlir
--- a/mlir/test/Conversion/TosaToLinalg/tosa-to-linalg.mlir
+++ b/mlir/test/Conversion/TosaToLinalg/tosa-to-linalg.mlir
@@ -258,3 +258,14 @@
   return
 }
 
+// -----
+
+func @test_identity(%arg0: tensor<1xf32>, %arg1: tensor<1xi32>) -> (tensor<1xf32>, tensor<1xi32>) {
+  %0 = "tosa.identity"(%arg0) : (tensor<1xf32>) -> tensor<1xf32>
+  %1 = "tosa.identity"(%arg1) : (tensor<1xi32>) -> tensor<1xi32>
+
+  %2:2 = "tosa.identityn"(%0, %1) : (tensor<1xf32>, tensor<1xi32>) -> (tensor<1xf32>, tensor<1xi32>)
+
+  // CHECK: return %arg0, %arg1
+  return %2#0, %2#1 : tensor<1xf32>, tensor<1xi32>
+}