Register fake_quantize related ops (llvm#3522)

Register `aten.fake_quantize_per_channel_affine` and `aten.fake_quantize_per_tensor_affine.tensor_qparams` ops --------- Co-authored-by: Ze Zhang <ze.zhang@getcruise.com>
knwng · Jul 5, 2024 · d466d5b · d466d5b
1 parent 0fe7484
commit d466d5b
Show file tree

Hide file tree

Showing 5 changed files with 149 additions and 1 deletion.
diff --git a/include/torch-mlir/Dialect/Torch/IR/GeneratedTorchOps.td b/include/torch-mlir/Dialect/Torch/IR/GeneratedTorchOps.td
@@ -4623,6 +4623,61 @@ def Torch_AtenFakeQuantizePerTensorAffineCachemaskOp : Torch_Op<"aten.fake_quant
   }];
 }
 
+def Torch_AtenFakeQuantizePerTensorAffineTensorQparamsOp : Torch_Op<"aten.fake_quantize_per_tensor_affine.tensor_qparams", [
+    AllowsTypeRefinement,
+    HasValueSemantics,
+    ReadOnly
+  ]> {
+  let summary = "Generated op for `aten::fake_quantize_per_tensor_affine.tensor_qparams : (Tensor, Tensor, Tensor, int, int) -> (Tensor)`";
+  let arguments = (ins
+    AnyTorchTensorType:$self,
+    AnyTorchTensorType:$scale,
+    AnyTorchTensorType:$zero_point,
+    Torch_IntType:$quant_min,
+    Torch_IntType:$quant_max
+  );
+  let results = (outs
+    AnyTorchOptionalTensorType:$result
+  );
+  let hasCustomAssemblyFormat = 1;
+  let extraClassDefinition = [{
+    ParseResult AtenFakeQuantizePerTensorAffineTensorQparamsOp::parse(OpAsmParser &parser, OperationState &result) {
+      return parseDefaultTorchOp(parser, result, 5, 1);
+    }
+    void AtenFakeQuantizePerTensorAffineTensorQparamsOp::print(OpAsmPrinter &printer) {
+      printDefaultTorchOp(printer, *this, 5, 1);
+    }
+  }];
+}
+
+def Torch_AtenFakeQuantizePerChannelAffineOp : Torch_Op<"aten.fake_quantize_per_channel_affine", [
+    AllowsTypeRefinement,
+    HasValueSemantics,
+    ReadOnly
+  ]> {
+  let summary = "Generated op for `aten::fake_quantize_per_channel_affine : (Tensor, Tensor, Tensor, int, int, int) -> (Tensor)`";
+  let arguments = (ins
+    AnyTorchTensorType:$self,
+    AnyTorchTensorType:$scale,
+    AnyTorchTensorType:$zero_point,
+    Torch_IntType:$axis,
+    Torch_IntType:$quant_min,
+    Torch_IntType:$quant_max
+  );
+  let results = (outs
+    AnyTorchOptionalTensorType:$result
+  );
+  let hasCustomAssemblyFormat = 1;
+  let extraClassDefinition = [{
+    ParseResult AtenFakeQuantizePerChannelAffineOp::parse(OpAsmParser &parser, OperationState &result) {
+      return parseDefaultTorchOp(parser, result, 6, 1);
+    }
+    void AtenFakeQuantizePerChannelAffineOp::print(OpAsmPrinter &printer) {
+      printDefaultTorchOp(printer, *this, 6, 1);
+    }
+  }];
+}
+
 def Torch_AtenMaximumOp : Torch_Op<"aten.maximum", [
     AllowsTypeRefinement,
     HasValueSemantics,

diff --git a/lib/Dialect/Torch/Transforms/AbstractInterpLibrary.cpp b/lib/Dialect/Torch/Transforms/AbstractInterpLibrary.cpp
@@ -6363,6 +6363,14 @@ StringRef mlir::torch::Torch::getAbstractInterpLibrary() {
 "    %2 = torch.prim.TupleConstruct %0, %1 : !torch.list<int>, !torch.list<int> -> !torch.tuple<list<int>, list<int>>\n"
 "    return %2 : !torch.tuple<list<int>, list<int>>\n"
 "  }\n"
+"  func.func @\"__torch_mlir_shape_fn.aten.fake_quantize_per_tensor_affine.tensor_qparams\"(%arg0: !torch.list<int>, %arg1: !torch.list<int>, %arg2: !torch.list<int>, %arg3: !torch.int, %arg4: !torch.int) -> !torch.list<int> {\n"
+"    %0 = call @__torch__.torch.jit._shape_functions.unary(%arg0) : (!torch.list<int>) -> !torch.list<int>\n"
+"    return %0 : !torch.list<int>\n"
+"  }\n"
+"  func.func @\"__torch_mlir_shape_fn.aten.fake_quantize_per_channel_affine\"(%arg0: !torch.list<int>, %arg1: !torch.list<int>, %arg2: !torch.list<int>, %arg3: !torch.int, %arg4: !torch.int, %arg5: !torch.int) -> !torch.list<int> {\n"
+"    %0 = call @__torch__.torch.jit._shape_functions.unary(%arg0) : (!torch.list<int>) -> !torch.list<int>\n"
+"    return %0 : !torch.list<int>\n"
+"  }\n"
 "  func.func @\"__torch_mlir_shape_fn.aten.sin\"(%arg0: !torch.list<int>) -> !torch.list<int> {\n"
 "    %0 = call @__torch__.torch.jit._shape_functions.unary(%arg0) : (!torch.list<int>) -> !torch.list<int>\n"
 "    return %0 : !torch.list<int>\n"
@@ -10551,6 +10559,48 @@ StringRef mlir::torch::Torch::getAbstractInterpLibrary() {
 "    %4 = torch.prim.TupleConstruct %3, %int11 : !torch.int, !torch.int -> !torch.tuple<int, int>\n"
 "    return %4 : !torch.tuple<int, int>\n"
 "  }\n"
+"  func.func @\"__torch_mlir_dtype_fn.aten.fake_quantize_per_tensor_affine.tensor_qparams\"(%arg0: !torch.tuple<int, int>, %arg1: !torch.tuple<int, int>, %arg2: !torch.tuple<int, int>, %arg3: !torch.int, %arg4: !torch.int) -> !torch.int {\n"
+"    %int15 = torch.constant.int 15\n"
+"    %none = torch.constant.none\n"
+"    %str = torch.constant.str \"AssertionError: \"\n"
+"    %0:2 = torch.prim.TupleUnpack %arg0 : !torch.tuple<int, int> -> !torch.int, !torch.int\n"
+"    %1 = call @__torch__.torch_mlir.jit_ir_importer.build_tools.library_generator.is_float_dtype(%0#1) : (!torch.int) -> !torch.bool\n"
+"    torch.prim.If %1 -> () {\n"
+"      torch.prim.If.yield\n"
+"    } else {\n"
+"      torch.prim.RaiseException %str, %none : !torch.str, !torch.none\n"
+"      torch.prim.If.yield\n"
+"    }\n"
+"    %2 = torch.aten.ne.int %0#1, %int15 : !torch.int, !torch.int -> !torch.bool\n"
+"    torch.prim.If %2 -> () {\n"
+"      torch.prim.If.yield\n"
+"    } else {\n"
+"      torch.prim.RaiseException %str, %none : !torch.str, !torch.none\n"
+"      torch.prim.If.yield\n"
+"    }\n"
+"    return %0#1 : !torch.int\n"
+"  }\n"
+"  func.func @\"__torch_mlir_dtype_fn.aten.fake_quantize_per_channel_affine\"(%arg0: !torch.tuple<int, int>, %arg1: !torch.tuple<int, int>, %arg2: !torch.tuple<int, int>, %arg3: !torch.int, %arg4: !torch.int, %arg5: !torch.int) -> !torch.int {\n"
+"    %int15 = torch.constant.int 15\n"
+"    %none = torch.constant.none\n"
+"    %str = torch.constant.str \"AssertionError: \"\n"
+"    %0:2 = torch.prim.TupleUnpack %arg0 : !torch.tuple<int, int> -> !torch.int, !torch.int\n"
+"    %1 = call @__torch__.torch_mlir.jit_ir_importer.build_tools.library_generator.is_float_dtype(%0#1) : (!torch.int) -> !torch.bool\n"
+"    torch.prim.If %1 -> () {\n"
+"      torch.prim.If.yield\n"
+"    } else {\n"
+"      torch.prim.RaiseException %str, %none : !torch.str, !torch.none\n"
+"      torch.prim.If.yield\n"
+"    }\n"
+"    %2 = torch.aten.ne.int %0#1, %int15 : !torch.int, !torch.int -> !torch.bool\n"
+"    torch.prim.If %2 -> () {\n"
+"      torch.prim.If.yield\n"
+"    } else {\n"
+"      torch.prim.RaiseException %str, %none : !torch.str, !torch.none\n"
+"      torch.prim.If.yield\n"
+"    }\n"
+"    return %0#1 : !torch.int\n"
+"  }\n"
 "  func.func @\"__torch_mlir_dtype_fn.aten.cosh\"(%arg0: !torch.tuple<int, int>) -> !torch.int {\n"
 "    %0:2 = torch.prim.TupleUnpack %arg0 : !torch.tuple<int, int> -> !torch.int, !torch.int\n"
 "    %1 = call @__torch__._get_dtype_of_floating_point_op(%0#1) : (!torch.int) -> !torch.int\n"

diff --git a/projects/pt1/python/torch_mlir/jit_ir_importer/build_tools/abstract_interp_lib_gen.py b/projects/pt1/python/torch_mlir/jit_ir_importer/build_tools/abstract_interp_lib_gen.py
@@ -135,6 +135,12 @@ def aten〇fake_quantize_per_tensor_affine〡shape(self: List[int], scale: float
 def aten〇fake_quantize_per_tensor_affine_cachemask〡shape(self: List[int], scale: float, zero_point: int, quant_min: int, quant_max: int) -> Tuple[List[int], List[int]]:
     return (upstream_shape_functions.unary(self), upstream_shape_functions.unary(self))
 
+def aten〇fake_quantize_per_tensor_affine〇tensor_qparams〡shape(self: List[int], scale: List[int], zero_point: List[int], quant_min: int, quant_max: int) -> List[int]:
+    return upstream_shape_functions.unary(self)
+
+def aten〇fake_quantize_per_channel_affine〡shape(self: List[int], scale: List[int], zero_point: List[int], axis: int, quant_min: int, quant_max: int) -> List[int]:
+    return upstream_shape_functions.unary(self)
+
 def aten〇sin〡shape(self: List[int]) -> List[int]:
     return upstream_shape_functions.unary(self)
 
@@ -2301,6 +2307,22 @@ def aten〇fake_quantize_per_tensor_affine_cachemask〡dtype(self_rank_dtype: Tu
     assert self_dtype != torch.bfloat16
     return (self_rank_dtype[1], torch.bool)
 
+# note: fake_quantize_per_tensor_affine.tensor_qparams doesn't support "meta" device, use "cpu" instead.
+@check_dtype_function(Invocation(TensorOfShape(3, 3, dtype=dtype, device="cpu"), TensorOfShape(1, dtype=torch.float32, device="cpu"), TensorOfShape(1, dtype=torch.int32, device="cpu"), 0, 255) for dtype in [torch.float64, torch.float32, torch.float16])
+def aten〇fake_quantize_per_tensor_affine〇tensor_qparams〡dtype(self_rank_dtype: Tuple[int, int], scale_rank_dtype: Tuple[int, int], zero_point_rank_dtype: Tuple[int, int], quant_min: int, quant_max: int) -> int:
+    self_rank, self_dtype = self_rank_dtype
+    assert is_float_dtype(self_dtype)
+    assert self_dtype != torch.bfloat16
+    return self_dtype
+
+# note: fake_quantize_per_channel_affine doesn't support "meta" device, use "cpu" instead.
+@check_dtype_function(Invocation(TensorOfShape(3, 3, dtype=dtype, device="cpu"), TensorOfShape(3, dtype=torch.float32, device="cpu"), TensorOfShape(3, dtype=torch.int32, device="cpu"), 0, 0, 255) for dtype in [torch.float64, torch.float32, torch.float16])
+def aten〇fake_quantize_per_channel_affine〡dtype(self_rank_dtype: Tuple[int, int], scale_rank_dtype: Tuple[int, int], zero_point_rank_dtype: Tuple[int, int], axis: int, quant_min: int, quant_max: int) -> int:
+    self_rank, self_dtype = self_rank_dtype
+    assert is_float_dtype(self_dtype)
+    assert self_dtype != torch.bfloat16
+    return self_dtype
+
 @check_dtype_function(_check_tensors_with_the_same_dtype(num_of_tensors=1))
 def aten〇cosh〡dtype(self_rank_dtype: Tuple[int, int]) -> int:
     self_rank, self_dtype = self_rank_dtype

diff --git a/projects/pt1/python/torch_mlir/jit_ir_importer/build_tools/torch_ods_gen.py b/projects/pt1/python/torch_mlir/jit_ir_importer/build_tools/torch_ods_gen.py
@@ -461,6 +461,12 @@ def emit_with_mutating_variants(key, **kwargs):
     emit(
         "aten::fake_quantize_per_tensor_affine_cachemask : (Tensor, float, int, int, int) -> (Tensor, Tensor)"
     )
+    emit(
+        "aten::fake_quantize_per_tensor_affine.tensor_qparams : (Tensor, Tensor, Tensor, int, int) -> (Tensor)"
+    )
+    emit(
+        "aten::fake_quantize_per_channel_affine : (Tensor, Tensor, Tensor, int, int, int) -> (Tensor)"
+    )
     emit("aten::maximum : (Tensor, Tensor) -> (Tensor)")
     emit("aten::minimum : (Tensor, Tensor) -> (Tensor)")
     emit("aten::fmax : (Tensor, Tensor) -> (Tensor)")

diff --git a/test/Dialect/Torch/ops.mlir b/test/Dialect/Torch/ops.mlir
@@ -188,5 +188,20 @@ func.func @torch.permute$negative_index_valid (%arg0: !torch.vtensor<[1,2,3],f32
   %int1 = torch.constant.int 1
   %perm = torch.prim.ListConstruct %int0, %int1, %intm1 : (!torch.int, !torch.int, !torch.int) -> !torch.list<int>
   %3 = torch.aten.permute %arg0, %perm : !torch.vtensor<[1,2,3],f32>, !torch.list<int> -> !torch.vtensor<[1,2,3],f32>
-   return %3 : !torch.vtensor<[1,2,3],f32>
+  return %3 : !torch.vtensor<[1,2,3],f32>
+}
+
+// Check fake quantize ops
+func.func @torch.aten.fake_quantize_per_channel_affine (%arg0: !torch.vtensor<[3,3],f32>, %arg1: !torch.vtensor<[3],f32>, %arg2: !torch.vtensor<[3],si32>) -> !torch.vtensor<[3,3],f32> {
+  %int0 = torch.constant.int 0
+  %int255 = torch.constant.int 255
+  %1 = torch.aten.fake_quantize_per_channel_affine %arg0, %arg1, %arg2, %int0, %int0, %int255 : !torch.vtensor<[3,3],f32>, !torch.vtensor<[3],f32>, !torch.vtensor<[3],si32>, !torch.int, !torch.int, !torch.int -> !torch.vtensor<[3,3],f32>
+  return %1 : !torch.vtensor<[3,3],f32>
+}
+
+func.func @torch.aten.fake_quantize_per_tensor_affine.tensor_qparams (%arg0: !torch.vtensor<[3,3],f32>, %arg1: !torch.vtensor<[3],f32>, %arg2: !torch.vtensor<[3],si32>) -> !torch.vtensor<[3,3],f32> {
+  %int0 = torch.constant.int 0
+  %int255 = torch.constant.int 255
+  %1 = torch.aten.fake_quantize_per_tensor_affine.tensor_qparams %arg0, %arg1, %arg2, %int0, %int255 : !torch.vtensor<[3,3],f32>, !torch.vtensor<[3],f32>, !torch.vtensor<[3],si32>, !torch.int, !torch.int -> !torch.vtensor<[3,3],f32>
+  return %1 : !torch.vtensor<[3,3],f32>
 }