rename namespace from dynamic to dyn

python/tvm/relay/__init__.py

@@ -45,7 +45,7 @@
 from .op import annotation
 from .op import vision
 from .op import contrib
-from .op import dynamic
+from .op import dyn
 from .op.reduce import *
 from .op.tensor import *
 from .op.transform import *

python/tvm/relay/op/dynamic/_make.py → python/tvm/relay/op/dyn/_make.py

@@ -17,4 +17,4 @@
 """Constructor APIs"""
 import tvm._ffi
 
-tvm._ffi._init_api("relay.op.dynamic._make", __name__)
+tvm._ffi._init_api("relay.op.dyn._make", __name__)

python/tvm/relay/op/dynamic/_transform.py → python/tvm/relay/op/dyn/_transform.py

@@ -20,7 +20,7 @@
 from tvm.te.hybrid import script
 from .. import op as _reg
 
-_reg.register_injective_schedule("dynamic.reshape")
+_reg.register_injective_schedule("dyn.reshape")
 
 @script
 def _reshape_shape_func_input_data(data, newshape, ndim):
@@ -78,6 +78,6 @@ def _reshape_shape_func_input_data(data, newshape, ndim):
             out[infer_idx] = old_size // new_size
     return out
 
-@_reg.register_shape_func("dynamic.reshape", True)
+@_reg.register_shape_func("dyn.reshape", True)
 def dynamic_reshape_shape_func(attrs, inputs, out_ndims):
     return [_reshape_shape_func_input_data(*inputs, out_ndims[0])]

src/relay/op/dynamic/tensor/transform.cc → src/relay/op/dyn/tensor/transform.cc

@@ -31,7 +31,7 @@
 
 namespace tvm {
 namespace relay {
-namespace dynamic {
+namespace dyn {
 
 /* relay.reshape */
 // TVM_REGISTER_NODE_TYPE(ReshapeAttrs);
@@ -78,13 +78,13 @@ Array<te::Tensor> ReshapeCompute(const Attrs& attrs, const Array<te::Tensor>& in
 Expr MakeReshape(Expr data, Expr newshape) {
   auto attrs = make_object<ReshapeAttrs>();
   attrs->reverse = false;
-  static const Op& op = Op::Get("dynamic.reshape");
+  static const Op& op = Op::Get("dyn.reshape");
   return Call(op, {data, newshape}, Attrs(attrs), {});
 }
 
-TVM_REGISTER_GLOBAL("relay.op.dynamic._make.reshape").set_body_typed(MakeReshape);
+TVM_REGISTER_GLOBAL("relay.op.dyn._make.reshape").set_body_typed(MakeReshape);
 
-RELAY_REGISTER_OP("dynamic.reshape")
+RELAY_REGISTER_OP("dyn.reshape")
     .describe(R"code(Reshapes the input array based on the values in the newshape array.
     
     To give user more convenience in without doing manual shape inference,
@@ -131,6 +131,6 @@ RELAY_REGISTER_OP("dynamic.reshape")
     .set_attr<FTVMCompute>("FTVMCompute", ReshapeCompute)
     .set_attr<TOpPattern>("TOpPattern", kInjective);
 
-}  // namespace dynamic
+}  // namespace dyn
 }  // namespace relay
 }  // namespace tvm

src/relay/op/dynamic/tensor/transform.h → src/relay/op/dyn/tensor/transform.h

@@ -26,7 +26,7 @@
 
 namespace tvm {
 namespace relay {
-namespace dynamic {}  // namespace dynamic
+namespace dyn {}  // namespace dynamic
 }  // namespace relay
 }  // namespace tvm
 #endif  // TVM_RELAY_OP_DYNAMIC_TENSOR_TRANSFORM_H_

src/relay/transforms/dynamic_to_static.cc

@@ -32,7 +32,7 @@ namespace relay {
 
 class DynamicToStaticMutator : public MixedModeMutator {
  public:
-  DynamicToStaticMutator() : dyn_reshape_op_(Op::Get("dynamic.reshape")) {}
+  DynamicToStaticMutator() : dyn_reshape_op_(Op::Get("dyn.reshape")) {}
 
  private:
   Expr Rewrite_(const CallNode* pre, const Expr& post) override {

tests/python/relay/dynamic/test_dynamic_op_level3.py → tests/python/relay/dyn/test_dynamic_op_level3.py

@@ -36,11 +36,11 @@ def verify_func(func, data, ref_res):
             tvm.testing.assert_allclose(op_res.asnumpy(), ref_res, rtol=1e-5)
             relay.backend.compile_engine.get().clear()
 
-def test_dynamic_reshape():
+def test_dyn_reshape():
     def verify_reshape(shape, newshape, oshape):
         x = relay.var("x", relay.TensorType(shape, "float32"))
         y = relay.var("y", relay.TensorType((len(newshape), ), "int64"))
-        z = relay.dynamic.reshape(x, y)
+        z = relay.dyn.reshape(x, y)
 
         func = relay.Function([x, y], z)
         x_data = np.random.uniform(low=-1, high=1, size=shape).astype("float32")
@@ -56,11 +56,11 @@ def verify_reshape(shape, newshape, oshape):
     verify_reshape((2, 3, 4, 5), (-3, -3), (6, 20))
     verify_reshape((2, 3, 4), (0, -3), (2, 12))
 
-def test_dynamic_shape_reshape():
+def test_dyn_shape_reshape():
     def verify_reshape(shape, newshape, oshape):
         x = relay.var("x", relay.TensorType(shape, "float32"))
         y = relay.var("y", relay.TensorType(newshape, "float32"))
-        z = relay.dynamic.reshape(x, relay.shape_of(y))
+        z = relay.dyn.reshape(x, relay.shape_of(y))
 
         func = relay.Function([x, y], z)
         x_data = np.random.uniform(low=-1, high=1, size=shape).astype("float32")
@@ -71,5 +71,5 @@ def verify_reshape(shape, newshape, oshape):
     verify_reshape((4, 7), (2, 7, 2), (2, 7, 2))
 
 if __name__ == "__main__":
-    test_dynamic_reshape()
-    test_dynamic_shape_reshape()
+    test_dyn_reshape()
+    test_dyn_shape_reshape()

tests/python/relay/test_pass_dynamic_to_static.py

@@ -34,8 +34,6 @@ def run_opt_pass(expr, opt_pass):
 def verify_func(func, data, ref_res):
     assert isinstance(data, list)
     for target, ctx in ctx_list():
-        #TODO(mbrookhart): enable Cuda tests onces the VM supports dynamic shapes
-        if "llvm" not in target: continue
         for kind in ["graph", "vm", "debug"]:
             mod = tvm.ir.IRModule.from_expr(func)
             intrp = relay.create_executor(kind, mod=mod, ctx=ctx, target=target)
@@ -46,7 +44,7 @@ def test_dynamic_to_static_reshape():
     def verify_reshape(shape, newshape, oshape):
         x = relay.var("x", relay.TensorType(shape, "float32"))
         y = relay.var("y", relay.TensorType(newshape, "float32"))
-        z = relay.dynamic.reshape(x, relay.shape_of(y))
+        z = relay.dyn.reshape(x, relay.shape_of(y))
         func = run_infer_type(relay.Function([x, y], z))
         func2 = run_opt_pass(run_opt_pass(func, transform.DynamicToStatic()), transform.InferType())
 
@@ -68,8 +66,8 @@ def test_dynamic_to_static_double_reshape():
     def verify_reshape(shape, newshape):
         x = relay.var("x", relay.TensorType(shape, "float32"))
         y = relay.var("y", relay.TensorType(newshape, "float32"))
-        z = relay.dynamic.reshape(x, relay.shape_of(y))
-        z = relay.dynamic.reshape(z, relay.shape_of(x))
+        z = relay.dyn.reshape(x, relay.shape_of(y))
+        z = relay.dyn.reshape(z, relay.shape_of(x))
         func = run_infer_type(relay.Function([x, y], z))
         func2 = run_opt_pass(run_opt_pass(func, transform.DynamicToStatic()), transform.InferType())
 
@@ -90,10 +88,10 @@ def test_dynamic_to_static_quad_reshape():
     def verify_reshape(shape, newshape):
         x = relay.var("x", relay.TensorType(shape, "float32"))
         y = relay.var("y", relay.TensorType(newshape, "float32"))
-        z1 = relay.dynamic.reshape(x, relay.shape_of(y))
-        z2 = relay.dynamic.reshape(z1, relay.shape_of(x))
-        z3 = relay.dynamic.reshape(z2, relay.shape_of(z1))
-        z4 = relay.dynamic.reshape(z3, relay.shape_of(z2))
+        z1 = relay.dyn.reshape(x, relay.shape_of(y))
+        z2 = relay.dyn.reshape(z1, relay.shape_of(x))
+        z3 = relay.dyn.reshape(z2, relay.shape_of(z1))
+        z4 = relay.dyn.reshape(z3, relay.shape_of(z2))
         func = run_infer_type(relay.Function([x, y], z4))
         func2 = run_opt_pass(run_opt_pass(func, transform.DynamicToStatic()), transform.InferType())