[MLIR][TORCH] Add E2E support for aten.randint.low op

Signed-Off By: Vivek Khandelwal<vivek@nod-labs.com>

include/torch-mlir/Dialect/Torch/IR/GeneratedTorchOps.td

@@ -3334,6 +3334,35 @@ def Torch_AtenBernoulli_FloatOp : Torch_Op<"aten.bernoulli_.float", [
   }];
 }
 
+def Torch_AtenRandintLowOp : Torch_Op<"aten.randint.low", [
+    AllowsTypeRefinement,
+    HasValueSemantics,
+    ReadOnly
+  ]> {
+  let summary = "Generated op for `aten::randint.low : (int, int, int[], int?, int?, Device?, bool?) -> (Tensor)`";
+  let arguments = (ins
+    Torch_IntType:$low,
+    Torch_IntType:$high,
+    AnyTorchListOfTorchIntType:$size,
+    AnyTorchOptionalIntType:$dtype,
+    AnyTorchOptionalIntType:$layout,
+    AnyTorchOptionalDeviceType:$device,
+    AnyTorchOptionalBoolType:$pin_memory
+  );
+  let results = (outs
+    AnyTorchTensorType:$result
+  );
+  let hasCustomAssemblyFormat = 1;
+  let extraClassDefinition = [{
+    ParseResult AtenRandintLowOp::parse(OpAsmParser &parser, OperationState &result) {
+      return parseDefaultTorchOp(parser, result, 7, 1);
+    }
+    void AtenRandintLowOp::print(OpAsmPrinter &printer) {
+      printDefaultTorchOp(printer, *this, 7, 1);
+    }
+  }];
+}
+
 def Torch_AtenBernoulliTensorOp : Torch_Op<"aten.bernoulli.Tensor", [
     AllowsTypeRefinement,
     HasValueSemantics,

lib/Dialect/Torch/Transforms/DecomposeComplexOps.cpp

@@ -241,7 +241,7 @@ public:
         rewriter.create<ConstantIntOp>(loc, rewriter.getI64IntegerAttr(1));
     Value startPlusLength =
         rewriter.create<AtenAddIntOp>(loc, one.getType(), start, length);
-    
+
     rewriter.replaceOpWithNewOp<AtenSliceTensorOp>(
         op, op.getResult().getType(), op.self(), /*dim=*/dim, /*start=*/start,
         /*end=*/startPlusLength, /*step=*/one);
@@ -3063,6 +3063,56 @@ public:
 };
 } // namespace
 
+namespace {
+class DecomposeAtenRandintLowOp : public OpRewritePattern<AtenRandintLowOp> {
+public:
+  using OpRewritePattern::OpRewritePattern;
+  LogicalResult matchAndRewrite(AtenRandintLowOp op,
+                                PatternRewriter &rewriter) const override {
+
+    Location loc = op.getLoc();
+    Type resultType = op.getType();
+    BaseTensorType resultTensorType = resultType.cast<BaseTensorType>();
+
+    int64_t cstLow, cstHigh;
+    if (!matchPattern(op.low(), m_TorchConstantInt(&cstLow)))
+      return rewriter.notifyMatchFailure(
+          op, "unimplemented: low must be a constant integer");
+    if (!matchPattern(op.high(), m_TorchConstantInt(&cstHigh)))
+      return rewriter.notifyMatchFailure(
+          op, "unimplemented: high must be a constant integer");
+
+    Value none = rewriter.create<ConstantNoneOp>(loc);
+    Value cstFalse = rewriter.create<ConstantBoolOp>(loc, false);
+    Value low = rewriter.create<Torch::ConstantFloatOp>(
+        loc, rewriter.getF64FloatAttr((double)cstLow));
+    Value high = rewriter.create<Torch::ConstantFloatOp>(
+        loc, rewriter.getF64FloatAttr((double)cstHigh));
+
+    BaseTensorType floatResultType =
+        resultTensorType
+            .getWithSizesAndDtype(resultTensorType.getSizes(),
+                                  rewriter.getF32Type())
+            .cast<BaseTensorType>();
+    Value emptyTensor = rewriter.create<AtenEmptyMemoryFormatOp>(
+        loc, floatResultType, op.size(), /*dtype=*/none, /*layout=*/op.layout(),
+        /*device=*/op.device(), /*pin_memory=*/op.pin_memory(),
+        /*memory_format=*/none);
+
+    Value result =
+        rewriter.create<AtenUniformOp>(loc, floatResultType, emptyTensor,
+                                       /*from=*/low,
+                                       /*to=*/high,
+                                       /*generator=*/none);
+    rewriter.replaceOpWithNewOp<AtenToDtypeOp>(
+        op, resultType, result,
+        getDtypeIntValueForType(rewriter, loc, resultTensorType.getDtype()),
+        /*non_blocking=*/cstFalse, /*copy=*/cstFalse, /*memory_format=*/none);
+    return success();
+  }
+};
+} // namespace
+
 namespace {
 class DecomposeComplexOpsPass
     : public DecomposeComplexOpsBase<DecomposeComplexOpsPass> {
@@ -3264,6 +3314,8 @@ public:
     target.addIllegalOp<AtenIndexTensorHackedTwinOp>();
     patterns.add<DecomposeAtenMseLossOp>(context);
     target.addIllegalOp<AtenMseLossOp>();
+    patterns.add<DecomposeAtenRandintLowOp>(context);
+    target.addIllegalOp<AtenRandintLowOp>();
 
     for (std::string opName : legalOps) {
       target.addLegalOp(OperationName(opName, context));

lib/Dialect/Torch/Transforms/RefineTypes.cpp

@@ -1158,6 +1158,17 @@ void TypeAnalysis::visitOperation(Operation *op,
     return;
   }
 
+  if (auto randIntLow = dyn_cast<AtenRandintLowOp>(op)) {
+    auto knowledge =
+        ValueKnowledge::getTensorPessimisticValueState(op->getContext());
+    Type defaultDtype =
+        IntegerType::get(op->getContext(), 64, IntegerType::Signed);
+    knowledge.dtype =
+        getDtypeOrDefault(op->getContext(), randIntLow.dtype(), defaultDtype);
+    incorporateKnowledge(randIntLow.getResult(), knowledge);
+    return;
+  }
+
   // Otherwise, this is an unknown operation, so reset the state.
   setAllToEntryStates(results);
   return;

lib/Dialect/Torch/Transforms/ShapeLibrary.cpp

@@ -6327,6 +6327,9 @@ StringRef mlir::torch::Torch::getShapeLibrary() {
 "  func.func @\"__torch_mlir_shape_fn.aten.rand_like\"(%arg0: !torch.list<int>, %arg1: !torch.optional<int>, %arg2: !torch.optional<int>, %arg3: !torch.optional<Device>, %arg4: !torch.optional<bool>, %arg5: !torch.optional<int>) -> !torch.list<int> {\n"
 "    return %arg0 : !torch.list<int>\n"
 "  }\n"
+"  func.func @\"__torch_mlir_shape_fn.aten.randint.low\"(%arg0: !torch.int, %arg1: !torch.int, %arg2: !torch.list<int>, %arg3: !torch.optional<int>, %arg4: !torch.optional<int>, %arg5: !torch.optional<Device>, %arg6: !torch.optional<bool>) -> !torch.list<int> {\n"
+"    return %arg2 : !torch.list<int>\n"
+"  }\n"
 "  func.func @\"__torch_mlir_shape_fn.aten.arange.start_step\"(%arg0: !torch.float, %arg1: !torch.float, %arg2: !torch.float, %arg3: !torch.optional<int>, %arg4: !torch.optional<int>, %arg5: !torch.optional<Device>, %arg6: !torch.optional<bool>) -> !torch.list<int> {\n"
 "    %0 = torch.derefine %arg0 : !torch.float to !torch.union<float, int>\n"
 "    %1 = torch.derefine %arg1 : !torch.float to !torch.union<float, int>\n"

python/torch_mlir/dialects/torch/importer/jit_ir/build_tools/shape_lib_gen.py

@@ -828,6 +828,9 @@ def aten〇cumsum(self: List[int], dim: int, dtype: Optional[int] = None) -> Lis
 def aten〇rand_like(self: List[int], dtype: Optional[int] = None, layout: Optional[int] = None, device: Optional[device] = None, pin_memory: Optional[bool] = None, memory_format: Optional[int] = None) -> List[int]:
     return self
 
+def aten〇randint〇low(low: int, high: int, size: List[int], dtype: Optional[int] = 4, layout: Optional[int] = None, device: Optional[device] = None, pin_memory: Optional[bool] = None) -> List[int]:
+    return size
+
 def aten〇arange〇start_step(start: float, end: float, step: float = 1, dtype: Optional[int] = None, layout: Optional[int] = None, device: Optional[device] = None, pin_memory: Optional[bool] = None) -> List[int]:
     return upstream_shape_functions.arange_start_step(start, end, step, dtype, layout, device, pin_memory)
 

python/torch_mlir/dialects/torch/importer/jit_ir/build_tools/torch_ods_gen.py

@@ -324,6 +324,7 @@ def emit_ops(emitter_td: TextEmitter, registry: Registry):
     emit("aten::rand_like : (Tensor, int?, int?, Device?, bool?, int?) -> (Tensor)")
     emit("aten::bernoulli : (Tensor, Generator?) -> (Tensor)")
     emit("aten::bernoulli_.float : (Tensor, float, Generator?) -> (Tensor)")
+    emit("aten::randint.low : (int, int, int[], int?, int?, Device?, bool?) -> (Tensor)")
     emit_with_mutating_variants("aten::bernoulli.Tensor : (Tensor, Tensor, Generator?) -> (Tensor)")
 
     emit_with_mutating_variants("aten::triu : (Tensor, int) -> (Tensor)")

python/torch_mlir_e2e_test/test_suite/rng.py

@@ -276,3 +276,43 @@ class RandLikeDtypeModule(torch.nn.Module):
 @register_test_case(module_factory=lambda: RandLikeDtypeModule())
 def RandLikeDtypeModule_basic(module, tu: TestUtils):
     module.forward(tu.rand(1024, 1024).double())
+
+# ==============================================================================
+
+class RandIntLowModule(torch.nn.Module):
+    def __init__(self):
+        super().__init__()
+
+    @export
+    @annotate_args([
+        None,
+    ])
+    def forward(self):
+        a = torch.ops.aten.randint(low=1, high=1000, size=[1024, 1024])
+        mean = torch.mean(a.to(torch.float32))
+        return mean
+
+
+@register_test_case(module_factory=lambda: RandIntLowModule())
+def RandIntLowModule_basic(module, tu: TestUtils):
+    module.forward()
+
+# ==============================================================================
+
+class RandIntLowDtypeModule(torch.nn.Module):
+    def __init__(self):
+        super().__init__()
+
+    @export
+    @annotate_args([
+        None,
+    ])
+    def forward(self):
+        a = torch.ops.aten.randint(low=1, high=1000, size=[128, 256, 512], dtype=torch.float64)
+        mean = torch.mean(a)
+        return mean
+
+
+@register_test_case(module_factory=lambda: RandIntLowDtypeModule())
+def RandIntLowDtypeModule_basic(module, tu: TestUtils):
+    module.forward()