fix lint

Author: Matthew Brookhart

python/tvm/relay/frontend/onnx.py

@@ -2235,13 +2235,9 @@ def conditionally_squeeze_scalar(x):
         one = _op.const(np.array([1]), dtype="int64")
         two = _op.const(np.array([2]), dtype="int64")
         three = _op.const(np.array([3]), dtype="int64")
-        two_ones = _op.const(np.array([1, 1]), dtype="int64")
         three_ones = _op.const(np.array([1, 1, 1]), dtype="int64")
         four_ones = _op.const(np.array([1, 1, 1, 1]), dtype="int64")
 
-        def pad_last_dim(x):
-            return _op.expand_dims(x, -1, 1)
-
         # First Loop Vars
         i = _expr.var("i", shape=(1,), dtype="int64")
         scores_var = _expr.var("scores_var", shape=(_ty.Any(), _ty.Any(), _ty.Any()), dtype=dtype)

python/tvm/topi/cuda/nms.py

@@ -98,10 +98,6 @@ def get_valid_counts_ir(
     valid_count = ib.buffer_ptr(valid_count)
     out = ib.buffer_ptr(out)
     out_indices = ib.buffer_ptr(out_indices)
-    atomic_add_return = ib.allocate(
-        valid_count.dtype, (1,), name="atomic_add_return", scope="local"
-    )
-    one_count = tvm.tir.const(1, dtype=valid_count.dtype)
     one = tvm.tir.const(1, dtype=out.dtype)
     if isinstance(score_threshold, float):
         score_threshold = tvm.ir.make_node("FloatImm", dtype="float32", value=score_threshold)

src/relay/backend/vm/lambda_lift.cc

@@ -115,21 +115,21 @@ class LambdaLifter : public ExprMutator {
     Array<Var> typed_captured_vars;
     Map<Var, Expr> rebinding_map;
     for (auto free_var : captured_vars) {
-        auto var = Var(free_var->name_hint(), free_var->checked_type());
-        typed_captured_vars.push_back(var);
-        rebinding_map.Set(free_var, var);
+      auto var = Var(free_var->name_hint(), free_var->checked_type());
+      typed_captured_vars.push_back(var);
+      rebinding_map.Set(free_var, var);
     }
 
     if (recursive) {
       if (!captured_vars.empty()) {
-          Array<Expr> fvs;
-          for (auto fv : captured_vars) {
-            fvs.push_back(fv);
-          }
-          lambda_map_.emplace(letrec_.back(), Call(global, fvs));
-        } else {
-          lambda_map_.emplace(letrec_.back(), global);
+        Array<Expr> fvs;
+        for (auto fv : captured_vars) {
+          fvs.push_back(fv);
         }
+        lambda_map_.emplace(letrec_.back(), Call(global, fvs));
+      } else {
+        lambda_map_.emplace(letrec_.back(), global);
+      }
     }
 
     auto body = Downcast<Function>(ExprMutator::VisitExpr_(func_node));
@@ -173,10 +173,12 @@ class LambdaLifter : public ExprMutator {
       // construct the "closure" function with fully annotated arguments, no longer relying
       // on type inference.
       auto before = Downcast<Function>(body)->params.size();
-      auto rebound_body = Function(func->params, Bind(body->body, rebinding_map), func->ret_type, func->type_params, func->attrs, func->span);
+      auto rebound_body = Function(func->params, Bind(body->body, rebinding_map), func->ret_type,
+                                   func->type_params, func->attrs, func->span);
       auto after = Downcast<Function>(rebound_body)->params.size();
       CHECK_EQ(before, after);
-      lifted_func = Function(typed_captured_vars, rebound_body, func->func_type_annotation(), free_type_vars);
+      lifted_func =
+          Function(typed_captured_vars, rebound_body, func->func_type_annotation(), free_type_vars);
       lifted_func = MarkClosure(lifted_func);
     }
 

src/relay/op/tensor/transform.h

@@ -44,16 +44,15 @@ template <typename AttrType>
 bool ConcatenateRel(const Array<Type>& types, int num_inputs, const Attrs& attrs,
                     const TypeReporter& reporter) {
   // types: [data, result]
-  ICHECK_EQ(types.size(), 2)
-    << "the arity of concatenate is 2, not " << types.size();
+  ICHECK_EQ(types.size(), 2) << "the arity of concatenate is 2, not " << types.size();
   /* If we receive a tuple we can continue, if we receive
    * anything but an incomplete type we should signal an
    * error.
    */
   const auto* tensor_tuple = types[0].as<TupleTypeNode>();
   if (tensor_tuple == nullptr) {
     reporter->GetDiagCtx().EmitFatal(
-      Diagnostic::Error(reporter->GetSpan())
+        Diagnostic::Error(reporter->GetSpan())
         << "concatenate requires a tuple of tensors as the first argument, found "
         << PrettyPrint(types[0]));
     return false;
@@ -63,8 +62,8 @@ bool ConcatenateRel(const Array<Type>& types, int num_inputs, const Attrs& attrs
 
   const auto* param = attrs.as<AttrType>();
   if (param == nullptr) {
-    reporter->GetDiagCtx().EmitFatal(
-      Diagnostic::Error(reporter->GetSpan()) << "the call attributes are not defined");
+    reporter->GetDiagCtx().EmitFatal(Diagnostic::Error(reporter->GetSpan())
+                                     << "the call attributes are not defined");
     return false;
   }
 

tests/python/frontend/onnx/test_forward.py

@@ -55,6 +55,7 @@ def get_tvm_output_with_vm(
     )
 
     from tvm.relay import transform
+
     # print(mod.astext(show_meta_data=True))
     # self.mod = transform.AnnotateSpans()(mod)
     # print(mod.astext(show_meta_data=False))
@@ -3940,4 +3941,5 @@ def test_loop():
 if __name__ == "__main__":
     import sys
     import pytest
+
     pytest.main(sys.argv)