This is an automated email from the ASF dual-hosted git repository.
kparzysz pushed a commit to branch main
in repository https://gitbox.apache.org/repos/asf/tvm.git
The following commit(s) were added to refs/heads/main by this push:
new 3756b716d5 [Relay] Use f-strings for string formatting, NFC (#14838)
3756b716d5 is described below
commit 3756b716d537a158f668794555d5193e29100ad0
Author: Krzysztof Parzyszek <[email protected]>
AuthorDate: Sat May 13 09:20:21 2023 -0500
[Relay] Use f-strings for string formatting, NFC (#14838)
* [Relay] Use f-strings for string formatting, NFC
Replace uses of % and .format() with f-strings.
Reformat modified files.
* Fix typo in python/tvm/relay/frontend/tensorflow_ops.py
`0[s0_size] -> s0[s0_size]`
---
python/tvm/relay/backend/interpreter.py | 11 ++-
python/tvm/relay/backend/te_compiler.py | 4 +-
python/tvm/relay/build_module.py | 2 +-
python/tvm/relay/expr.py | 69 +++++---------
python/tvm/relay/expr_functor.py | 8 +-
python/tvm/relay/frontend/tensorflow_ops.py | 2 +-
python/tvm/relay/loops.py | 2 +-
python/tvm/relay/prelude.py | 76 +++++-----------
python/tvm/relay/qnn/op/layout_conversions.py | 4 +-
python/tvm/relay/qnn/op/qnn.py | 81 +++--------------
python/tvm/relay/quantize/_calibrate.py | 4 +-
python/tvm/relay/quantize/quantize.py | 4 +-
python/tvm/relay/testing/dcgan.py | 4 +-
python/tvm/relay/testing/densenet.py | 12 +--
python/tvm/relay/testing/inception_v3.py | 100 +++++++++------------
python/tvm/relay/testing/init.py | 11 ++-
python/tvm/relay/testing/layers.py | 12 +--
python/tvm/relay/testing/lstm.py | 20 ++---
python/tvm/relay/testing/mobilenet.py | 2 +-
python/tvm/relay/testing/py_converter.py | 21 ++---
python/tvm/relay/testing/resnet.py | 8 +-
python/tvm/relay/testing/resnet_3d.py | 8 +-
python/tvm/relay/testing/squeezenet.py | 17 ++--
python/tvm/relay/testing/tf.py | 8 +-
python/tvm/relay/testing/tflite.py | 2 +-
python/tvm/relay/testing/vgg.py | 8 +-
.../transform/fake_quantization_to_integer.py | 30 ++-----
python/tvm/relay/type_functor.py | 2 +-
28 files changed, 185 insertions(+), 347 deletions(-)
diff --git a/python/tvm/relay/backend/interpreter.py
b/python/tvm/relay/backend/interpreter.py
index e4da6f447f..80a8880fbc 100644
--- a/python/tvm/relay/backend/interpreter.py
+++ b/python/tvm/relay/backend/interpreter.py
@@ -99,7 +99,7 @@ class Executor(object):
if kwargs and not isinstance(expr, Function):
raise Exception(
- "can only supply keyword parameters for a " "relay.Function,
found {0}".format(expr)
+ f"can only supply keyword parameters for a relay.Function,
found {expr}"
)
params = expr.params
@@ -111,17 +111,16 @@ class Executor(object):
if i < num_of_args:
if kwargs.get(name):
raise Exception(
- "duplicate argument supplied in "
- "both positional args (at position: {0}), "
- "and keyword argument (with name: {1})".format(i, name)
+ f"duplicate argument supplied in "
+ f"both positional args (at position: {i}), "
+ f"and keyword argument (with name: {name})"
)
else:
cargs.append(kwargs[name])
if len(cargs) != len(params):
raise Exception(
- "insufficient arguments, expected "
- "{0}, provided {1}".format(len(cargs), len(params))
+ f"insufficient arguments, expected " f"{len(cargs)}, provided
{len(params)}"
)
return tuple(cargs)
diff --git a/python/tvm/relay/backend/te_compiler.py
b/python/tvm/relay/backend/te_compiler.py
index 814e793290..84e4ecbaec 100644
--- a/python/tvm/relay/backend/te_compiler.py
+++ b/python/tvm/relay/backend/te_compiler.py
@@ -111,8 +111,8 @@ def get_valid_implementations(op, attrs, inputs, out_type,
target):
"""
fstrategy = op.get_attr("FTVMStrategy")
assert fstrategy is not None, (
- "%s doesn't have an FTVMStrategy registered. You can register "
- "one in python with `tvm.relay.op.register_strategy`." % op.name
+ f"{op.name} doesn't have an FTVMStrategy registered. You can register "
+ f"one in python with `tvm.relay.op.register_strategy`."
)
with target:
strategy = fstrategy(attrs, inputs, out_type, target)
diff --git a/python/tvm/relay/build_module.py b/python/tvm/relay/build_module.py
index f2feed9fd6..40a91cc75a 100644
--- a/python/tvm/relay/build_module.py
+++ b/python/tvm/relay/build_module.py
@@ -683,4 +683,4 @@ def create_executor(kind="debug", mod=None, device=None,
target="llvm", params=N
return VMExecutor(mod, device, raw_targets)
if kind == "aot":
return AotExecutor(mod, device, raw_targets)
- raise RuntimeError("unknown execution strategy: {0}".format(kind))
+ raise RuntimeError(f"unknown execution strategy: {kind}")
diff --git a/python/tvm/relay/expr.py b/python/tvm/relay/expr.py
index d8bca5c4a4..5239eaa883 100644
--- a/python/tvm/relay/expr.py
+++ b/python/tvm/relay/expr.py
@@ -97,41 +97,41 @@ class ExprWithOp(RelayExpr):
if isinstance(other, Expr):
return _op_make.less(self, other)
elif isinstance(other, _Number):
- raise TypeError('convert "%s" with `const` first' % str(other))
+ raise TypeError(f'convert "{str(other)}" with `const` first')
else:
- raise TypeError("type %s not supported" % str(type(other)))
+ raise TypeError(f"type {type(other)} not supported")
def __gt__(self, other):
if isinstance(other, Expr):
return _op_make.greater(self, other)
elif isinstance(other, _Number):
- raise TypeError('convert "%s" with `const` first' % str(other))
+ raise TypeError(f'convert "{str(other)}" with `const` first')
else:
- raise TypeError("type %s not supported" % str(type(other)))
+ raise TypeError(f"type {type(other)} not supported")
def __ge__(self, other):
if isinstance(other, Expr):
return _op_make.greater_equal(self, other)
elif isinstance(other, _Number):
- raise TypeError('convert "%s" with `const` first' % str(other))
+ raise TypeError(f'convert "{str(other)}" with `const` first')
else:
- raise TypeError("type %s not supported" % str(type(other)))
+ raise TypeError(f"type {type(other)} not supported")
def __le__(self, other):
if isinstance(other, Expr):
return _op_make.less_equal(self, other)
elif isinstance(other, _Number):
- raise TypeError('convert "%s" with `const` first' % str(other))
+ raise TypeError(f'convert "{str(other)}" with `const` first')
else:
- raise TypeError("type %s not supported" % str(type(other)))
+ raise TypeError(f"type {type(other)} not supported")
def __add__(self, other):
if isinstance(other, Expr):
return _op_make.add(self, other)
elif isinstance(other, _Number):
- raise TypeError('convert "%s" with `const` first' % str(other))
+ raise TypeError(f'convert "{str(other)}" with `const` first')
else:
- raise TypeError("type %s not supported" % str(type(other)))
+ raise TypeError(f"type {type(other)} not supported")
def __radd__(self, other):
return self.__add__(other)
@@ -140,22 +140,22 @@ class ExprWithOp(RelayExpr):
if isinstance(other, Expr):
return _op_make.subtract(self, other)
elif isinstance(other, _Number):
- raise TypeError('convert "%s" with `const` first' % str(other))
+ raise TypeError(f'convert "{str(other)}" with `const` first')
else:
- raise TypeError("type %s not supported" % str(type(other)))
+ raise TypeError(f"type {type(other)} not supported")
def __rsub__(self, other):
if isinstance(other, _Number):
- raise TypeError('convert "%s" with `const` first' % str(other))
- raise TypeError("type %s not supported" % str(type(other)))
+ raise TypeError(f'convert "{str(other)}" with `const` first')
+ raise TypeError(f"type {type(other)} not supported")
def __mul__(self, other):
if isinstance(other, Expr):
return _op_make.multiply(self, other)
elif isinstance(other, _Number):
- raise TypeError('convert "%s" with `const` first' % str(other))
+ raise TypeError(f'convert "{str(other)}" with `const` first')
else:
- raise TypeError("type %s not supported" % str(type(other)))
+ raise TypeError(f"type {type(other)} not supported")
def __rmul__(self, other):
return self.__mul__(other)
@@ -164,14 +164,14 @@ class ExprWithOp(RelayExpr):
if isinstance(other, Expr):
return _op_make.divide(self, other)
elif isinstance(other, _Number):
- raise TypeError('convert "%s" with `const` first' % str(other))
+ raise TypeError(f'convert "{str(other)}" with `const` first')
else:
- raise TypeError("type %s not supported" % str(type(other)))
+ raise TypeError(f"type {type(other)} not supported")
def __rdiv__(self, other):
if isinstance(other, _Number):
- raise TypeError('convert "%s" with `const` first' % str(other))
- raise TypeError("type %s not supported" % str(type(other)))
+ raise TypeError(f'convert "{str(other)}" with `const` first')
+ raise TypeError(f"type {type(other)} not supported")
def __truediv__(self, other):
return self.__div__(other)
@@ -213,12 +213,7 @@ class Constant(ExprWithOp):
@tvm._ffi.register_func("relay.ConstantWithFields")
-def ConstantWithFields(
- constant,
- data=None,
- virtual_device=None,
- span=None,
-):
+def ConstantWithFields(constant, data=None, virtual_device=None, span=None):
"""
Returns constant with the given properties. A None property denotes 'no
change'.
Returns constant if all properties are unchanged. Otherwise, returns a
copy with the new
@@ -467,12 +462,7 @@ class RefCreate(ExprWithOp):
@tvm._ffi.register_func("relay.RefCreateWithFields")
-def RefCreateWithFields(
- ref_create,
- value=None,
- virtual_device=None,
- span=None,
-):
+def RefCreateWithFields(ref_create, value=None, virtual_device=None,
span=None):
"""
Returns ref_create with the given properties. A None property denotes 'no
change'.
Returns ref_create if all properties are unchanged. Otherwise, returns a
copy with the new
@@ -498,12 +488,7 @@ class RefRead(ExprWithOp):
@tvm._ffi.register_func("relay.RefReadWithFields")
-def RefReadWithFields(
- ref_read,
- ref=None,
- virtual_device=None,
- span=None,
-):
+def RefReadWithFields(ref_read, ref=None, virtual_device=None, span=None):
"""
Returns ref_read with the given properties. A None property denotes 'no
change'.
Returns ref_read if all properties are unchanged. Otherwise, returns a
copy with the new
@@ -534,13 +519,7 @@ class RefWrite(ExprWithOp):
@tvm._ffi.register_func("relay.RefWriteWithFields")
-def RefWriteWithFields(
- ref_write,
- ref=None,
- value=None,
- virtual_device=None,
- span=None,
-):
+def RefWriteWithFields(ref_write, ref=None, value=None, virtual_device=None,
span=None):
"""
Returns ref_write with the given properties. A None property denotes 'no
change'.
Returns ref_write if all properties are unchanged. Otherwise, returns a
copy with the new
diff --git a/python/tvm/relay/expr_functor.py b/python/tvm/relay/expr_functor.py
index ebea344b41..95a8c79dc2 100644
--- a/python/tvm/relay/expr_functor.py
+++ b/python/tvm/relay/expr_functor.py
@@ -73,7 +73,7 @@ class ExprFunctor:
elif isinstance(expr, Match):
res = self.visit_match(expr)
else:
- raise Exception("warning unhandled case: {0}".format(type(expr)))
+ raise Exception(f"warning unhandled case: {type(expr)}")
self.memo_map[expr] = res
@@ -204,11 +204,7 @@ class ExprMutator(ExprFunctor):
def visit_function(self, fn):
new_params = [self.visit(x) for x in fn.params]
new_body = self.visit(fn.body)
- return FunctionWithFields(
- fn,
- list(new_params),
- new_body,
- )
+ return FunctionWithFields(fn, list(new_params), new_body)
def visit_let(self, let):
new_var = self.visit(let.var)
diff --git a/python/tvm/relay/frontend/tensorflow_ops.py
b/python/tvm/relay/frontend/tensorflow_ops.py
index 014d0065fc..e2c3a34252 100644
--- a/python/tvm/relay/frontend/tensorflow_ops.py
+++ b/python/tvm/relay/frontend/tensorflow_ops.py
@@ -1946,7 +1946,7 @@ def _broadcast_args():
else:
assert (
s1[s1_size - i] == 1
- ), f"Incompatible broadcast type {0[s0_size - i]} and
{s1[s1_size - i]}"
+ ), f"Incompatible broadcast type {s0[s0_size - i]} and
{s1[s1_size - i]}"
out.appendleft(s0[s0_size - i])
if s0_size < s1_size:
for i in range(s0_size + 1, s1_size + 1):
diff --git a/python/tvm/relay/loops.py b/python/tvm/relay/loops.py
index d46e34860f..61183fd531 100644
--- a/python/tvm/relay/loops.py
+++ b/python/tvm/relay/loops.py
@@ -53,7 +53,7 @@ def while_loop(cond, loop_vars, loop_bodies):
fresh_vars = []
for i, loop_var in enumerate(loop_vars):
- name = loop_var.name_hint if isinstance(loop_var, _expr.Var) else
"arg{}".format(i)
+ name = loop_var.name_hint if isinstance(loop_var, _expr.Var) else
f"arg{i}"
new_var = _expr.var(name, type_annotation=sb.type_of(loop_var),
span=loop_var.span)
fresh_vars.append(new_var)
diff --git a/python/tvm/relay/prelude.py b/python/tvm/relay/prelude.py
index f21e3eaf2c..0db639a3a8 100644
--- a/python/tvm/relay/prelude.py
+++ b/python/tvm/relay/prelude.py
@@ -59,9 +59,9 @@ def get_tensor_array_shape(expr, dtype, prelude):
checked_type = mod["main"].body.checked_type
assert isinstance(checked_type, TypeCall), "Input must be a tensor array."
ta_type_str = checked_type.args[0].func.name_hint
- static_ta_ty_start = "static_tensor_{}".format(dtype)
+ static_ta_ty_start = f"static_tensor_{dtype}"
if ta_type_str.startswith(static_ta_ty_start):
- shape_str = ta_type_str.replace("{}_".format(static_ta_ty_start),
"").replace("_t", "")
+ shape_str = ta_type_str.replace(f"{static_ta_ty_start}_",
"").replace("_t", "")
shape = []
if "scalar" not in shape_str:
for dim_str in shape_str.split("_"):
@@ -104,18 +104,18 @@ def _get_name_static(canonical, dtype, shape,
batch_dim=None, extra_shapes=None)
if extra_shapes is not None:
for n, s in extra_shapes.items():
- extra_shape_str = "_{}_{}".format(n, _to_str(s))
+ extra_shape_str = f"_{n}_{_to_str(s)}"
shape_str += extra_shape_str
if len(shape_str) == 0:
shape_str = "scalar"
if canonical == "tensor_t":
- return "static_tensor_{}_{}_t".format(dtype, shape_str)
+ return f"static_tensor_{dtype}_{shape_str}_t"
if batch_dim is None or canonical in ["tensor_constructor", "tensor_nil"]:
- return "{}_{}_{}".format(canonical, dtype, shape_str)
+ return f"{canonical}_{dtype}_{shape_str}"
if batch_dim != 1:
- return "{}_{}_{}".format(canonical, dtype, shape_str)
- return "{}_{}_batch{}_{}".format(canonical, dtype, str(batch_dim),
shape_str)
+ return f"{canonical}_{dtype}_{shape_str}"
+ return f"{canonical}_{dtype}_batch{batch_dim}_{shape_str}"
def _to_str(shape):
@@ -224,9 +224,7 @@ class StaticTensorArrayOps(object):
origin_tensor_constructor = self.get_ctor("tensor_constructor")
- output_shape = [
- Any(),
- ] + list(self.shape[1:])
+ output_shape = [Any()] + list(self.shape[1:])
tensor_type_var, tensor_constructor, _ =
self._get_adt_by_shape(output_shape)
t = Var("tensor", self.tensor_type_var())
@@ -255,9 +253,7 @@ class StaticTensorArrayOps(object):
if self.is_cached(concat_name):
return
- output_shape = [
- Any(),
- ] + list(self.shape[1:])
+ output_shape = [Any()] + list(self.shape[1:])
tensor_type_var, tensor_constructor, _ =
self._get_adt_by_shape(output_shape)
origin_tensor_constructor = self.get_ctor("tensor_constructor")
@@ -301,10 +297,7 @@ class StaticTensorArrayOps(object):
# in stack op, we need to recursively concatenate.
new_axis = Any() if self.batch_dim is None or self.batch_dim != 1 else
self.batch_dim
tensor_type_var, tensor_constructor, _ = self._get_adt_by_shape(
- [
- new_axis,
- ]
- + list(self.shape)
+ [new_axis] + list(self.shape)
)
t = Var("t")
case = Clause(
@@ -497,9 +490,7 @@ class StaticTensorArrayOps(object):
tensor_array_split_helper_var =
GlobalVar(tensor_array_split_helper_name)
split_var = GlobalVar(split_name)
- output_shape = [
- Any(),
- ] + list(self.shape[1:])
+ output_shape = [Any()] + list(self.shape[1:])
output_tensor_type_var, _, output_ops =
self._get_adt_by_shape(output_shape)
output_ops.define_tensor_array_write()
write_var = output_ops.get_global_var("tensor_array_write")
@@ -575,9 +566,7 @@ class StaticTensorArrayOps(object):
concat_var = GlobalVar(concat_name)
- output_shape = [
- Any(),
- ] + list(self.shape[1:])
+ output_shape = [Any()] + list(self.shape[1:])
tensor_type_var, _, output_ops = self._get_adt_by_shape(output_shape)
@@ -617,9 +606,7 @@ class StaticTensorArrayOps(object):
# Register tensor_concatenate for output_shape
new_axis = Any() if not self.batch_dim or self.batch_dim != 1 else
self.batch_dim
- output_shape = [
- new_axis,
- ] + list(self.shape)
+ output_shape = [new_axis] + list(self.shape)
_, _, output_ops = self._get_adt_by_shape(output_shape)
output_ops.define_tensor_concatenate()
concat_var = output_ops.get_global_var("tensor_concatenate")
@@ -627,9 +614,7 @@ class StaticTensorArrayOps(object):
tensor_array_expand_dims = self.prelude.map(expand_dims_var,
tensor_array)
if self.batch_dim is not None and self.batch_dim == 1:
# only one element
- tensors = self.prelude.id(
- self.prelude.hd(tensor_array_expand_dims),
- )
+ tensors =
self.prelude.id(self.prelude.hd(tensor_array_expand_dims))
else:
tensors = self.prelude.foldl(
concat_var,
@@ -650,9 +635,7 @@ class StaticTensorArrayOps(object):
helper_var = self._create_global_var(helper_name)
new_axis = Any() if self.batch_dim is None or self.batch_dim != 1 else
self.batch_dim
- output_shape = [
- new_axis,
- ] + list(self.shape)
+ output_shape = [new_axis] + list(self.shape)
output_tensor_type_var, _, _ = self._get_adt_by_shape(output_shape)
stack_var = self.get_global_var("tensor_array_stack")
read_var = self.get_global_var("tensor_array_read")
@@ -1130,10 +1113,7 @@ class TensorArrayOps(object):
shape = op.shape_of(tensor2)
ndim = op.take(shape, const(0))
self.prelude.mod[tensor_array_unstack_tensor2_var] = Function(
- [tensor2],
- helper_var(const(0), ndim, tensor2),
- self.list(self.tensor_type_var()),
- [],
+ [tensor2], helper_var(const(0), ndim, tensor2),
self.list(self.tensor_type_var()), []
)
def define_tensor_array_unstack_tensor3(self):
@@ -1167,10 +1147,7 @@ class TensorArrayOps(object):
shape = op.shape_of(tensor3)
ndim = op.take(shape, const(0))
self.prelude.mod[tensor_array_unstack_tensor3_var] = Function(
- [tensor3],
- helper_var(const(0), ndim, tensor3),
- self.list(self.tensor_type_var()),
- [],
+ [tensor3], helper_var(const(0), ndim, tensor3),
self.list(self.tensor_type_var()), []
)
def define_tensor_array_unstack_tensor4(self):
@@ -1204,10 +1181,7 @@ class TensorArrayOps(object):
shape = op.shape_of(tensor4)
ndim = op.take(shape, const(0))
self.prelude.mod[tensor_array_unstack_tensor4_var] = Function(
- [tensor4],
- helper_var(const(0), ndim, tensor4),
- self.list(self.tensor_type_var()),
- [],
+ [tensor4], helper_var(const(0), ndim, tensor4),
self.list(self.tensor_type_var()), []
)
def define_tensor_array_unstack_tensor5(self):
@@ -1241,10 +1215,7 @@ class TensorArrayOps(object):
shape = op.shape_of(tensor5)
ndim = op.take(shape, const(0))
self.prelude.mod[tensor_array_unstack_tensor5_var] = Function(
- [tensor5],
- helper_var(const(0), ndim, tensor5),
- self.list(self.tensor_type_var()),
- [],
+ [tensor5], helper_var(const(0), ndim, tensor5),
self.list(self.tensor_type_var()), []
)
def define_tensor_array_unstack_tensor6(self):
@@ -1278,10 +1249,7 @@ class TensorArrayOps(object):
shape = op.shape_of(tensor6)
ndim = op.take(shape, const(0))
self.prelude.mod[tensor_array_unstack_tensor6_var] = Function(
- [tensor6],
- helper_var(const(0), ndim, tensor6),
- self.list(self.tensor_type_var()),
- [],
+ [tensor6], helper_var(const(0), ndim, tensor6),
self.list(self.tensor_type_var()), []
)
def define_tensor_array_scatter(self):
@@ -1507,8 +1475,8 @@ class Prelude:
def get_name(self, canonical, dtype):
"""Get name corresponding to the canonical name"""
if canonical == "tensor_t":
- return "tensor_{}_t".format(dtype)
- return "{}_{}".format(canonical, dtype)
+ return f"tensor_{dtype}_t"
+ return f"{canonical}_{dtype}"
def get_global_var(self, canonical, dtype):
"""Get global var corresponding to the canonical name"""
diff --git a/python/tvm/relay/qnn/op/layout_conversions.py
b/python/tvm/relay/qnn/op/layout_conversions.py
index 24c787e0a0..668cafb8ae 100644
--- a/python/tvm/relay/qnn/op/layout_conversions.py
+++ b/python/tvm/relay/qnn/op/layout_conversions.py
@@ -77,7 +77,7 @@ def convert_qnn_conv2d(attrs, inputs, tinfos,
desired_layouts):
new_attrs["kernel_layout"] = "HWIO"
return relay.qnn.op.conv2d(*inputs, **new_attrs)
- raise ValueError("Layout %s is not yet supported" % desired_data_layout)
+ raise ValueError(f"Layout {desired_data_layout} is not yet supported")
@reg.register_convert_op_layout("qnn.conv2d_transpose")
@@ -125,4 +125,4 @@ def convert_qnn_conv2d_transpose(attrs, inputs, tinfos,
desired_layouts):
new_attrs["kernel_layout"] = "HWIO"
return relay.qnn.op.conv2d_transpose(*inputs, **new_attrs)
- raise ValueError("Layout %s is not yet supported" % desired_data_layout)
+ raise ValueError(f"Layout {desired_data_layout} is not yet supported")
diff --git a/python/tvm/relay/qnn/op/qnn.py b/python/tvm/relay/qnn/op/qnn.py
index 504688759e..e2c251ec78 100644
--- a/python/tvm/relay/qnn/op/qnn.py
+++ b/python/tvm/relay/qnn/op/qnn.py
@@ -86,7 +86,7 @@ class RequantizeConfig(Object):
def __setattr__(self, name, value):
if name in RequantizeConfig._node_defaults:
- raise AttributeError("'%s' object cannot set attribute '%s'" %
(str(type(self)), name))
+ raise AttributeError(f"'{type(self)}' object cannot set attribute
'{name}'")
return super(RequantizeConfig, self).__setattr__(name, value)
@@ -876,13 +876,7 @@ def tanh(x, scale, zero_point, output_scale,
output_zero_point):
The computed result.
"""
- return _make.tanh(
- x,
- scale,
- zero_point,
- output_scale,
- output_zero_point,
- )
+ return _make.tanh(x, scale, zero_point, output_scale, output_zero_point)
def exp(x, scale, zero_point, output_scale, output_zero_point):
@@ -911,13 +905,7 @@ def exp(x, scale, zero_point, output_scale,
output_zero_point):
The computed result.
"""
- return _make.exp(
- x,
- scale,
- zero_point,
- output_scale,
- output_zero_point,
- )
+ return _make.exp(x, scale, zero_point, output_scale, output_zero_point)
def sqrt(x, scale, zero_point, output_scale, output_zero_point):
@@ -946,13 +934,7 @@ def sqrt(x, scale, zero_point, output_scale,
output_zero_point):
The computed result.
"""
- return _make.sqrt(
- x,
- scale,
- zero_point,
- output_scale,
- output_zero_point,
- )
+ return _make.sqrt(x, scale, zero_point, output_scale, output_zero_point)
def rsqrt(x, scale, zero_point, output_scale, output_zero_point):
@@ -981,13 +963,7 @@ def rsqrt(x, scale, zero_point, output_scale,
output_zero_point):
The computed result.
"""
- return _make.rsqrt(
- x,
- scale,
- zero_point,
- output_scale,
- output_zero_point,
- )
+ return _make.rsqrt(x, scale, zero_point, output_scale, output_zero_point)
def erf(x, scale, zero_point, output_scale, output_zero_point):
@@ -1016,13 +992,7 @@ def erf(x, scale, zero_point, output_scale,
output_zero_point):
The computed result.
"""
- return _make.erf(
- x,
- scale,
- zero_point,
- output_scale,
- output_zero_point,
- )
+ return _make.erf(x, scale, zero_point, output_scale, output_zero_point)
# pylint: disable=redefined-builtin
@@ -1054,13 +1024,7 @@ def abs(x, scale, zero_point, output_scale,
output_zero_point):
The computed result.
"""
- return _make.abs(
- x,
- scale,
- zero_point,
- output_scale,
- output_zero_point,
- )
+ return _make.abs(x, scale, zero_point, output_scale, output_zero_point)
def sigmoid(x, scale, zero_point, output_scale, output_zero_point):
@@ -1089,13 +1053,7 @@ def sigmoid(x, scale, zero_point, output_scale,
output_zero_point):
The computed result.
"""
- return _make.sigmoid(
- x,
- scale,
- zero_point,
- output_scale,
- output_zero_point,
- )
+ return _make.sigmoid(x, scale, zero_point, output_scale, output_zero_point)
def hardswish(x, scale, zero_point, output_scale, output_zero_point):
@@ -1124,13 +1082,7 @@ def hardswish(x, scale, zero_point, output_scale,
output_zero_point):
The computed result.
"""
- return _make.hardswish(
- x,
- scale,
- zero_point,
- output_scale,
- output_zero_point,
- )
+ return _make.hardswish(x, scale, zero_point, output_scale,
output_zero_point)
def log(x, scale, zero_point, output_scale, output_zero_point):
@@ -1159,13 +1111,7 @@ def log(x, scale, zero_point, output_scale,
output_zero_point):
The computed result.
"""
- return _make.log(
- x,
- scale,
- zero_point,
- output_scale,
- output_zero_point,
- )
+ return _make.log(x, scale, zero_point, output_scale, output_zero_point)
def subtract(
@@ -1297,10 +1243,5 @@ def leaky_relu(x, alpha, input_scale, input_zero_point,
output_scale, output_zer
The computed result.
"""
return _make.leaky_relu(
- x,
- alpha,
- input_scale,
- input_zero_point,
- output_scale,
- output_zero_point,
+ x, alpha, input_scale, input_zero_point, output_scale,
output_zero_point
)
diff --git a/python/tvm/relay/quantize/_calibrate.py
b/python/tvm/relay/quantize/_calibrate.py
index 4b2d55ebe8..f03d556814 100644
--- a/python/tvm/relay/quantize/_calibrate.py
+++ b/python/tvm/relay/quantize/_calibrate.py
@@ -224,14 +224,14 @@ def calibrate(dataset=None):
elif cfg.calibrate_mode == "percentile":
input_scale_func = _percentile_scale(mod, dataset)
else:
- raise ValueError("Unknown calibrate mode
{}".format(cfg.calibrate_mode))
+ raise ValueError(f"Unknown calibrate mode {cfg.calibrate_mode}")
if cfg.weight_scale == "max":
weight_scale_func = _max_scale
elif cfg.weight_scale == "power2":
weight_scale_func = _power2_scale
else:
- raise ValueError("Unknown weight scale mode
{}".format(cfg.weight_scale))
+ raise ValueError(f"Unknown weight scale mode {cfg.weight_scale}")
return _set_params(mod, input_scale_func, weight_scale_func)
diff --git a/python/tvm/relay/quantize/quantize.py
b/python/tvm/relay/quantize/quantize.py
index 7f4724db22..41343061da 100644
--- a/python/tvm/relay/quantize/quantize.py
+++ b/python/tvm/relay/quantize/quantize.py
@@ -128,7 +128,7 @@ class QConfig(Object):
def __setattr__(self, name, value):
if name in QConfig._node_defaults:
- raise AttributeError("'%s' object cannot set attribute '%s'" %
(str(type(self)), name))
+ raise AttributeError(f"'{type(self)}' object cannot set attribute
'{name}'")
return super(QConfig, self).__setattr__(name, value)
@@ -304,7 +304,7 @@ def _bind_params(func, params):
continue
arg = name_dict[k]
if arg is None:
- raise ValueError("Multiple args in the function have name %s" % k)
+ raise ValueError(f"Multiple args in the function have name {k}")
bind_dict[arg] = _expr.const(v)
return _expr.bind(func, bind_dict)
diff --git a/python/tvm/relay/testing/dcgan.py
b/python/tvm/relay/testing/dcgan.py
index acc478330d..4749d76dbc 100644
--- a/python/tvm/relay/testing/dcgan.py
+++ b/python/tvm/relay/testing/dcgan.py
@@ -65,10 +65,10 @@ def deconv2d(data, ishape, oshape, kshape, layout, name,
stride=(2, 2)):
def deconv2d_bn_relu(data, prefix, **kwargs):
"""a block of deconv + batch norm + relu"""
eps = 1e-5 + 1e-12
- net = deconv2d(data, name="%s_deconv" % prefix, **kwargs)
+ net = deconv2d(data, name=f"{prefix}_deconv", **kwargs)
bn_axis = kwargs.get("layout", "NCHW").index("C")
net = layers.batch_norm_infer(
- net, epsilon=eps, scale=False, axis=bn_axis, name="%s_batch_norm" %
prefix
+ net, epsilon=eps, scale=False, axis=bn_axis,
name=f"{prefix}_batch_norm"
)
net = relay.nn.relu(net)
return net
diff --git a/python/tvm/relay/testing/densenet.py
b/python/tvm/relay/testing/densenet.py
index 6b8d0098a5..c9deb78683 100644
--- a/python/tvm/relay/testing/densenet.py
+++ b/python/tvm/relay/testing/densenet.py
@@ -28,15 +28,15 @@ from .init import create_workload
def _make_dense_layer(data, growth_rate, bn_size, index):
"""Single densenet layer."""
- bn1 = layers.batch_norm_infer(data, name="batch_1_%s" % index)
+ bn1 = layers.batch_norm_infer(data, name=f"batch_1_{index}")
relu1 = relay.nn.relu(bn1)
conv1 = layers.conv2d(
- relu1, channels=bn_size * growth_rate, kernel_size=(1, 1),
name="conv2d_1_%s" % index
+ relu1, channels=bn_size * growth_rate, kernel_size=(1, 1),
name=f"conv2d_1_{index}"
)
bn2 = layers.batch_norm_infer(conv1, name="batch_2_" + index)
relu2 = relay.nn.relu(bn2)
conv2 = layers.conv2d(
- relu2, channels=growth_rate, kernel_size=(3, 3), padding=(1, 1),
name="conv2d_2_%s" % index
+ relu2, channels=growth_rate, kernel_size=(3, 3), padding=(1, 1),
name=f"conv2d_2_{index}"
)
return conv2
@@ -46,7 +46,7 @@ def _make_dense_block(data, num_layers, bn_size, growth_rate,
index):
layer_out = data
blocks = []
for i in range(num_layers):
- layer_out = _make_dense_layer(layer_out, growth_rate, bn_size, "%s_%s"
% (index, i))
+ layer_out = _make_dense_layer(layer_out, growth_rate, bn_size,
f"{index}_{i}")
blocks.append(layer_out)
block_out = relay.concatenate(blocks, 1)
return block_out
@@ -54,10 +54,10 @@ def _make_dense_block(data, num_layers, bn_size,
growth_rate, index):
def _make_transition(data, num_output_features, index):
"""Transition between layers."""
- bn = layers.batch_norm_infer(data, name="batch_t_%s" % index)
+ bn = layers.batch_norm_infer(data, name=f"batch_t_{index}")
relu = relay.nn.relu(bn)
conv = layers.conv2d(
- relu, channels=num_output_features, kernel_size=(1, 1),
name="conv_t_%s" % index
+ relu, channels=num_output_features, kernel_size=(1, 1),
name=f"conv_t_{index}"
)
return relay.nn.avg_pool2d(conv, pool_size=(2, 2), strides=(2, 2))
diff --git a/python/tvm/relay/testing/inception_v3.py
b/python/tvm/relay/testing/inception_v3.py
index 2381551f66..e5b89ccdec 100644
--- a/python/tvm/relay/testing/inception_v3.py
+++ b/python/tvm/relay/testing/inception_v3.py
@@ -37,12 +37,10 @@ def Conv(data, num_filter, kernel=(1, 1), stride=(1, 1),
pad=(0, 0), name=None,
kernel_size=kernel,
strides=stride,
padding=pad,
- name="%s%s_conv1" % (name, suffix),
+ name=f"{name}{suffix}_conv1",
)
- bn = layers.batch_norm_infer(
- data=conv, epsilon=2e-5, scale=False, name="%s%s_bn" % (name, suffix)
- )
+ bn = layers.batch_norm_infer(data=conv, epsilon=2e-5, scale=False,
name=f"{name}{suffix}_bn")
act = relay.nn.relu(data=bn)
return act
@@ -60,27 +58,17 @@ def Pooling(data, kernel, stride, pad, pool_type, name):
def Inception7A(
data, num_1x1, num_3x3_red, num_3x3_1, num_3x3_2, num_5x5_red, num_5x5,
pool, proj, name
):
- tower_1x1 = Conv(data, num_1x1, name=("%s_conv" % name))
- tower_5x5 = Conv(data, num_5x5_red, name=("%s_tower" % name),
suffix="_conv")
+ tower_1x1 = Conv(data, num_1x1, name=f"{name}_conv")
+ tower_5x5 = Conv(data, num_5x5_red, name=f"{name}_tower", suffix="_conv")
tower_5x5 = Conv(
- tower_5x5, num_5x5, kernel=(5, 5), pad=(2, 2), name=("%s_tower" %
name), suffix="_conv_1"
+ tower_5x5, num_5x5, kernel=(5, 5), pad=(2, 2), name=f"{name}_tower",
suffix="_conv_1"
)
- tower_3x3 = Conv(data, num_3x3_red, name=("%s_tower_1" % name),
suffix="_conv")
+ tower_3x3 = Conv(data, num_3x3_red, name=f"{name}_tower_1", suffix="_conv")
tower_3x3 = Conv(
- tower_3x3,
- num_3x3_1,
- kernel=(3, 3),
- pad=(1, 1),
- name=("%s_tower_1" % name),
- suffix="_conv_1",
+ tower_3x3, num_3x3_1, kernel=(3, 3), pad=(1, 1),
name=f"{name}_tower_1", suffix="_conv_1"
)
tower_3x3 = Conv(
- tower_3x3,
- num_3x3_2,
- kernel=(3, 3),
- pad=(1, 1),
- name=("%s_tower_1" % name),
- suffix="_conv_2",
+ tower_3x3, num_3x3_2, kernel=(3, 3), pad=(1, 1),
name=f"{name}_tower_1", suffix="_conv_2"
)
pooling = Pooling(
data=data,
@@ -88,27 +76,25 @@ def Inception7A(
stride=(1, 1),
pad=(1, 1),
pool_type=pool,
- name=("%s_pool_%s_pool" % (pool, name)),
+ name=f"{pool}_pool_{name}_pool",
)
- cproj = Conv(pooling, proj, name=("%s_tower_2" % name), suffix="_conv")
+ cproj = Conv(pooling, proj, name=f"{name}_tower_2", suffix="_conv")
concat = relay.concatenate((tower_1x1, tower_5x5, tower_3x3, cproj),
axis=1)
return concat
# First Downsample
def Inception7B(data, num_3x3, num_d3x3_red, num_d3x3_1, num_d3x3_2, pool,
name):
- tower_3x3 = Conv(
- data, num_3x3, kernel=(3, 3), pad=(0, 0), stride=(2, 2),
name=("%s_conv" % name)
- )
- tower_d3x3 = Conv(data, num_d3x3_red, name=("%s_tower" % name),
suffix="_conv")
+ tower_3x3 = Conv(data, num_3x3, kernel=(3, 3), pad=(0, 0), stride=(2, 2),
name=f"{name}_conv")
+ tower_d3x3 = Conv(data, num_d3x3_red, name=f"{name}_tower", suffix="_conv")
tower_d3x3 = Conv(
tower_d3x3,
num_d3x3_1,
kernel=(3, 3),
pad=(1, 1),
stride=(1, 1),
- name=("%s_tower" % name),
+ name=f"{name}_tower",
suffix="_conv_1",
)
tower_d3x3 = Conv(
@@ -117,7 +103,7 @@ def Inception7B(data, num_3x3, num_d3x3_red, num_d3x3_1,
num_d3x3_2, pool, name)
kernel=(3, 3),
pad=(0, 0),
stride=(2, 2),
- name=("%s_tower" % name),
+ name=f"{name}_tower",
suffix="_conv_2",
)
pooling = Pooling(
@@ -126,7 +112,7 @@ def Inception7B(data, num_3x3, num_d3x3_red, num_d3x3_1,
num_d3x3_2, pool, name)
stride=(2, 2),
pad=(0, 0),
pool_type="max",
- name=("max_pool_%s_pool" % name),
+ name=f"max_pool_{name}_pool",
)
concat = relay.concatenate((tower_3x3, tower_d3x3, pooling), axis=1)
return concat
@@ -147,14 +133,14 @@ def Inception7C(
proj,
name,
):
- tower_1x1 = Conv(data=data, num_filter=num_1x1, kernel=(1, 1),
name=("%s_conv" % name))
- tower_d7 = Conv(data=data, num_filter=num_d7_red, name=("%s_tower" %
name), suffix="_conv")
+ tower_1x1 = Conv(data=data, num_filter=num_1x1, kernel=(1, 1),
name=f"{name}_conv")
+ tower_d7 = Conv(data=data, num_filter=num_d7_red, name=f"{name}_tower",
suffix="_conv")
tower_d7 = Conv(
data=tower_d7,
num_filter=num_d7_1,
kernel=(1, 7),
pad=(0, 3),
- name=("%s_tower" % name),
+ name=f"{name}_tower",
suffix="_conv_1",
)
tower_d7 = Conv(
@@ -162,16 +148,16 @@ def Inception7C(
num_filter=num_d7_2,
kernel=(7, 1),
pad=(3, 0),
- name=("%s_tower" % name),
+ name=f"{name}_tower",
suffix="_conv_2",
)
- tower_q7 = Conv(data=data, num_filter=num_q7_red, name=("%s_tower_1" %
name), suffix="_conv")
+ tower_q7 = Conv(data=data, num_filter=num_q7_red, name=f"{name}_tower_1",
suffix="_conv")
tower_q7 = Conv(
data=tower_q7,
num_filter=num_q7_1,
kernel=(7, 1),
pad=(3, 0),
- name=("%s_tower_1" % name),
+ name=f"{name}_tower_1",
suffix="_conv_1",
)
tower_q7 = Conv(
@@ -179,7 +165,7 @@ def Inception7C(
num_filter=num_q7_2,
kernel=(1, 7),
pad=(0, 3),
- name=("%s_tower_1" % name),
+ name=f"{name}_tower_1",
suffix="_conv_2",
)
tower_q7 = Conv(
@@ -187,7 +173,7 @@ def Inception7C(
num_filter=num_q7_3,
kernel=(7, 1),
pad=(3, 0),
- name=("%s_tower_1" % name),
+ name=f"{name}_tower_1",
suffix="_conv_3",
)
tower_q7 = Conv(
@@ -195,7 +181,7 @@ def Inception7C(
num_filter=num_q7_4,
kernel=(1, 7),
pad=(0, 3),
- name=("%s_tower_1" % name),
+ name=f"{name}_tower_1",
suffix="_conv_4",
)
pooling = Pooling(
@@ -204,10 +190,10 @@ def Inception7C(
stride=(1, 1),
pad=(1, 1),
pool_type=pool,
- name=("%s_pool_%s_pool" % (pool, name)),
+ name=f"{pool}_pool_{name}_pool",
)
cproj = Conv(
- data=pooling, num_filter=proj, kernel=(1, 1), name=("%s_tower_2" %
name), suffix="_conv"
+ data=pooling, num_filter=proj, kernel=(1, 1), name=f"{name}_tower_2",
suffix="_conv"
)
# concat
concat = relay.concatenate((tower_1x1, tower_d7, tower_q7, cproj), axis=1)
@@ -217,25 +203,25 @@ def Inception7C(
def Inception7D(
data, num_3x3_red, num_3x3, num_d7_3x3_red, num_d7_1, num_d7_2,
num_d7_3x3, pool, name
):
- tower_3x3 = Conv(data=data, num_filter=num_3x3_red, name=("%s_tower" %
name), suffix="_conv")
+ tower_3x3 = Conv(data=data, num_filter=num_3x3_red, name=f"{name}_tower",
suffix="_conv")
tower_3x3 = Conv(
data=tower_3x3,
num_filter=num_3x3,
kernel=(3, 3),
pad=(0, 0),
stride=(2, 2),
- name=("%s_tower" % name),
+ name=f"{name}_tower",
suffix="_conv_1",
)
tower_d7_3x3 = Conv(
- data=data, num_filter=num_d7_3x3_red, name=("%s_tower_1" % name),
suffix="_conv"
+ data=data, num_filter=num_d7_3x3_red, name=f"{name}_tower_1",
suffix="_conv"
)
tower_d7_3x3 = Conv(
data=tower_d7_3x3,
num_filter=num_d7_1,
kernel=(1, 7),
pad=(0, 3),
- name=("%s_tower_1" % name),
+ name=f"{name}_tower_1",
suffix="_conv_1",
)
tower_d7_3x3 = Conv(
@@ -243,7 +229,7 @@ def Inception7D(
num_filter=num_d7_2,
kernel=(7, 1),
pad=(3, 0),
- name=("%s_tower_1" % name),
+ name=f"{name}_tower_1",
suffix="_conv_2",
)
tower_d7_3x3 = Conv(
@@ -251,7 +237,7 @@ def Inception7D(
num_filter=num_d7_3x3,
kernel=(3, 3),
stride=(2, 2),
- name=("%s_tower_1" % name),
+ name=f"{name}_tower_1",
suffix="_conv_3",
)
pooling = Pooling(
@@ -260,7 +246,7 @@ def Inception7D(
stride=(2, 2),
pool_type=pool,
pad=(0, 0),
- name=("%s_pool_%s_pool" % (pool, name)),
+ name=f"{pool}_pool_{name}_pool",
)
# concat
concat = relay.concatenate((tower_3x3, tower_d7_3x3, pooling), axis=1)
@@ -281,14 +267,14 @@ def Inception7E(
proj,
name,
):
- tower_1x1 = Conv(data=data, num_filter=num_1x1, kernel=(1, 1),
name=("%s_conv" % name))
- tower_d3 = Conv(data=data, num_filter=num_d3_red, name=("%s_tower" %
name), suffix="_conv")
+ tower_1x1 = Conv(data=data, num_filter=num_1x1, kernel=(1, 1),
name=f"{name}_conv")
+ tower_d3 = Conv(data=data, num_filter=num_d3_red, name=f"{name}_tower",
suffix="_conv")
tower_d3_a = Conv(
data=tower_d3,
num_filter=num_d3_1,
kernel=(1, 3),
pad=(0, 1),
- name=("%s_tower" % name),
+ name=f"{name}_tower",
suffix="_mixed_conv",
)
tower_d3_b = Conv(
@@ -296,18 +282,18 @@ def Inception7E(
num_filter=num_d3_2,
kernel=(3, 1),
pad=(1, 0),
- name=("%s_tower" % name),
+ name=f"{name}_tower",
suffix="_mixed_conv_1",
)
tower_3x3_d3 = Conv(
- data=data, num_filter=num_3x3_d3_red, name=("%s_tower_1" % name),
suffix="_conv"
+ data=data, num_filter=num_3x3_d3_red, name=f"{name}_tower_1",
suffix="_conv"
)
tower_3x3_d3 = Conv(
data=tower_3x3_d3,
num_filter=num_3x3,
kernel=(3, 3),
pad=(1, 1),
- name=("%s_tower_1" % name),
+ name=f"{name}_tower_1",
suffix="_conv_1",
)
tower_3x3_d3_a = Conv(
@@ -315,7 +301,7 @@ def Inception7E(
num_filter=num_3x3_d3_1,
kernel=(1, 3),
pad=(0, 1),
- name=("%s_tower_1" % name),
+ name=f"{name}_tower_1",
suffix="_mixed_conv",
)
tower_3x3_d3_b = Conv(
@@ -323,7 +309,7 @@ def Inception7E(
num_filter=num_3x3_d3_2,
kernel=(3, 1),
pad=(1, 0),
- name=("%s_tower_1" % name),
+ name=f"{name}_tower_1",
suffix="_mixed_conv_1",
)
pooling = Pooling(
@@ -332,10 +318,10 @@ def Inception7E(
stride=(1, 1),
pad=(1, 1),
pool_type=pool,
- name=("%s_pool_%s_pool" % (pool, name)),
+ name=f"{pool}_pool_{name}_pool",
)
cproj = Conv(
- data=pooling, num_filter=proj, kernel=(1, 1), name=("%s_tower_2" %
name), suffix="_conv"
+ data=pooling, num_filter=proj, kernel=(1, 1), name=f"{name}_tower_2",
suffix="_conv"
)
# concat
concat = relay.concatenate(
diff --git a/python/tvm/relay/testing/init.py b/python/tvm/relay/testing/init.py
index f275712c77..373b5a8ec3 100644
--- a/python/tvm/relay/testing/init.py
+++ b/python/tvm/relay/testing/init.py
@@ -75,10 +75,10 @@ class Initializer(object):
def _init_default(self, name, _):
raise ValueError(
- "Unknown initialization pattern for %s. "
- "Default initialization is now limited to "
- '"weight", "bias", "gamma" (1.0), and "beta" (0.0).'
- "Please use mx.sym.Variable(init=mx.init.*) to set initialization
pattern" % name
+ f"Unknown initialization pattern for {name}. "
+ f"Default initialization is now limited to "
+ f'"weight", "bias", "gamma" (1.0), and "beta" (0.0).'
+ f"Please use mx.sym.Variable(init=mx.init.*) to set initialization
pattern"
)
@@ -110,8 +110,7 @@ class Xavier(Initializer):
hw_scale = 1.0
if len(shape) < 2:
raise ValueError(
- "Xavier initializer cannot be applied to vector {0}. It
requires at"
- " least 2D.".format(name)
+ f"Xavier initializer cannot be applied to vector {name}. It
requires at least 2D."
)
if len(shape) > 2:
hw_scale = np.prod(shape[2:])
diff --git a/python/tvm/relay/testing/layers.py
b/python/tvm/relay/testing/layers.py
index 48003f2ae2..8496c56400 100644
--- a/python/tvm/relay/testing/layers.py
+++ b/python/tvm/relay/testing/layers.py
@@ -189,14 +189,8 @@ def conv_kernel_layout(data_layout, is_depthwise=False):
result : str
The corresponding kernel layout.
"""
- conv_layout_map = {
- "NCHW": "OIHW",
- "NHWC": "HWIO",
- }
- depthwise_conv_layout_map = {
- "NCHW": "OIHW",
- "NHWC": "HWOI",
- }
+ conv_layout_map = {"NCHW": "OIHW", "NHWC": "HWIO"}
+ depthwise_conv_layout_map = {"NCHW": "OIHW", "NHWC": "HWOI"}
mapping = depthwise_conv_layout_map if is_depthwise else conv_layout_map
- assert data_layout in mapping, "Unknown data layout %s" % data_layout
+ assert data_layout in mapping, f"Unknown data layout {data_layout}"
return mapping[data_layout]
diff --git a/python/tvm/relay/testing/lstm.py b/python/tvm/relay/testing/lstm.py
index 8a97c18a1f..bf054592b0 100644
--- a/python/tvm/relay/testing/lstm.py
+++ b/python/tvm/relay/testing/lstm.py
@@ -69,7 +69,7 @@ def lstm_cell(num_hidden, batch_size=1, dtype="float32",
name=""):
i2h = builder.let(
("i2h", dense_type),
layers.dense_add_bias(
- data=inputs, units=num_hidden * 4, weight=i2h_weight,
bias=i2h_bias, name="%si2h" % name
+ data=inputs, units=num_hidden * 4, weight=i2h_weight,
bias=i2h_bias, name=f"{name}i2h"
),
)
h2h = builder.let(
@@ -79,7 +79,7 @@ def lstm_cell(num_hidden, batch_size=1, dtype="float32",
name=""):
units=num_hidden * 4,
weight=h2h_weight,
bias=h2h_bias,
- name="%sh2h" % name,
+ name=f"{name}h2h",
),
)
@@ -138,19 +138,19 @@ def get_net(iterations, num_hidden, batch_size=1,
dtype="float32"):
for i in range(iterations):
inputs = relay.Var("data", input_type)
- i2h_weight = relay.Var("i2h_%s_weight" % i, weight_type)
- i2h_bias = relay.Var("i2h_%i_bias" % i, bias_type)
- h2h_weight = relay.Var("h2h_%s_weight" % i, weight_type)
- h2h_bias = relay.Var("h2h_%s_bias" % i, bias_type)
+ i2h_weight = relay.Var(f"i2h_{i}_weight", weight_type)
+ i2h_bias = relay.Var(f"i2h_{i}_bias", bias_type)
+ h2h_weight = relay.Var(f"h2h_{i}_weight", weight_type)
+ h2h_bias = relay.Var(f"h2h_{i}_bias", bias_type)
- cell_fn = lstm_cell(num_hidden, batch_size, dtype, "lstm_%s" % i)
+ cell_fn = lstm_cell(num_hidden, batch_size, dtype, f"lstm_{i}")
call = builder.let(
- ("call_%s" % i, cell_type),
+ (f"call_{i}", cell_type),
relay.Call(cell_fn, [inputs, states, i2h_weight, i2h_bias,
h2h_weight, h2h_bias]),
)
- new_out = builder.let(("out_%s" % i, input_type),
relay.TupleGetItem(call, 0))
- new_states = builder.let(("states_%s" % i, state_type),
relay.TupleGetItem(call, 1))
+ new_out = builder.let((f"out_{i}", input_type),
relay.TupleGetItem(call, 0))
+ new_states = builder.let((f"states_{i}", state_type),
relay.TupleGetItem(call, 1))
states = new_states
out = new_out
diff --git a/python/tvm/relay/testing/mobilenet.py
b/python/tvm/relay/testing/mobilenet.py
index 0b5593eedc..4c600966d2 100644
--- a/python/tvm/relay/testing/mobilenet.py
+++ b/python/tvm/relay/testing/mobilenet.py
@@ -188,7 +188,7 @@ def mobile_net(
for i in range(7, 12):
body = separable_conv_block(
body,
- "separable_conv_block_%d" % i,
+ f"separable_conv_block_{i}",
int(512 * alpha),
int(512 * alpha),
layout=layout,
diff --git a/python/tvm/relay/testing/py_converter.py
b/python/tvm/relay/testing/py_converter.py
index 44489aa9cf..9cbfcead47 100644
--- a/python/tvm/relay/testing/py_converter.py
+++ b/python/tvm/relay/testing/py_converter.py
@@ -133,13 +133,13 @@ class PythonConverter(ExprFunctor):
def generate_var_name(self, name_hint: str) -> str:
"""Generates a unique variable name starting from the hint."""
- name = "{}_var_{}".format(self.sanitize(name_hint), self.var_no)
+ name = f"{self.sanitize(name_hint)}_var_{self.var_no}"
self.var_no += 1
return name
def generate_function_name(self, name_hint: str) -> str:
"""Generates a unique function name starting from the hint."""
- name = "{}_fun_{}".format(self.sanitize(name_hint), self.fun_no)
+ name = f"{self.sanitize(name_hint)}_fun_{self.fun_no}"
self.fun_no += 1
return name
@@ -261,11 +261,7 @@ class PythonConverter(ExprFunctor):
arguments = ast.arguments(inner_args, None, [], [], None, [])
return ast.FunctionDef(
- func_name,
- arguments,
- body,
- decorator_list if register_packed else [],
- None,
+ func_name, arguments, body, decorator_list if register_packed else
[], None
)
def create_tuple(self, fields):
@@ -285,7 +281,7 @@ class PythonConverter(ExprFunctor):
# compile the function and register globally
cc_key = te_compiler.CCacheKey(op, self.tgt)
func_hash = tvm.ir.structural_hash(op)
- op_name = "_lowered_op_{}".format(func_hash)
+ op_name = f"_lowered_op_{func_hash}"
if not tvm.get_global_func(op_name, allow_missing=True):
jitted = self.tec.jit(cc_key, self.tgt)
tvm.register_func(op_name, jitted)
@@ -334,8 +330,8 @@ class PythonConverter(ExprFunctor):
# create a function to wrap the call of the lowered op and return
# a call to that function
- wrap_name = self.generate_function_name("_{}_wrapper".format(op_name))
- wrap_args = [self.generate_var_name("_arg_{}".format(i)) for i in
range(len(py_args))]
+ wrap_name = self.generate_function_name(f"_{op_name}_wrapper")
+ wrap_args = [self.generate_var_name(f"_arg_{i}") for i in
range(len(py_args))]
inner_call_args = []
for i in range(len(py_args)):
@@ -588,10 +584,7 @@ class PythonConverter(ExprFunctor):
[],
ref_defs
+ val_defs
- + [
- Assign([ast.Attribute(ref, "value", Store())], val),
- Return(self.create_tuple([])),
- ],
+ + [Assign([ast.Attribute(ref, "value", Store())], val),
Return(self.create_tuple([]))],
)
return (self.create_call(thunk_name, []), [thunk])
diff --git a/python/tvm/relay/testing/resnet.py
b/python/tvm/relay/testing/resnet.py
index b35e01f677..e1e4069f54 100644
--- a/python/tvm/relay/testing/resnet.py
+++ b/python/tvm/relay/testing/resnet.py
@@ -239,7 +239,7 @@ def resnet(
filter_list[i + 1],
(1 if i == 0 else 2, 1 if i == 0 else 2),
False,
- name="stage%d_unit%d" % (i + 1, 1),
+ name=f"stage{i + 1}_unit1",
bottle_neck=bottle_neck,
data_layout=data_layout,
kernel_layout=kernel_layout,
@@ -250,7 +250,7 @@ def resnet(
filter_list[i + 1],
(1, 1),
True,
- name="stage%d_unit%d" % (i + 1, j + 2),
+ name=f"stage{i + 1}_unit{j + 2}",
bottle_neck=bottle_neck,
data_layout=data_layout,
kernel_layout=kernel_layout,
@@ -293,7 +293,7 @@ def get_net(
filter_list = [16, 16, 32, 64]
bottle_neck = False
else:
- raise ValueError("no experiments done on num_layers
{}".format(num_layers))
+ raise ValueError(f"no experiments done on num_layers {num_layers}")
units = per_unit * num_stages
else:
if num_layers >= 50:
@@ -318,7 +318,7 @@ def get_net(
elif num_layers == 269:
units = [3, 30, 48, 8]
else:
- raise ValueError("no experiments done on num_layers
{}".format(num_layers))
+ raise ValueError(f"no experiments done on num_layers {num_layers}")
return resnet(
units=units,
diff --git a/python/tvm/relay/testing/resnet_3d.py
b/python/tvm/relay/testing/resnet_3d.py
index 715e3951b8..b20833402a 100644
--- a/python/tvm/relay/testing/resnet_3d.py
+++ b/python/tvm/relay/testing/resnet_3d.py
@@ -233,7 +233,7 @@ def resnet(
filter_list[i + 1],
(1 if i == 0 else 2, 1 if i == 0 else 2, 1 if i == 0 else 2),
False,
- name="stage%d_unit%d" % (i + 1, 1),
+ name=f"stage{i + 1}_unit1",
bottle_neck=bottle_neck,
data_layout=data_layout,
kernel_layout=kernel_layout,
@@ -244,7 +244,7 @@ def resnet(
filter_list[i + 1],
(1, 1, 1),
True,
- name="stage%d_unit%d" % (i + 1, j + 2),
+ name=f"stage{i + 1}_unit{j + 2}",
bottle_neck=bottle_neck,
data_layout=data_layout,
kernel_layout=kernel_layout,
@@ -288,7 +288,7 @@ def get_net(
filter_list = [16, 16, 32, 64]
bottle_neck = False
else:
- raise ValueError("no experiments done on num_layers
{}".format(num_layers))
+ raise ValueError(f"no experiments done on num_layers {num_layers}")
units = per_unit * num_stages
else:
if num_layers >= 50:
@@ -313,7 +313,7 @@ def get_net(
elif num_layers == 269:
units = [3, 30, 48, 8]
else:
- raise ValueError("no experiments done on num_layers
{}".format(num_layers))
+ raise ValueError(f"no experiments done on num_layers {num_layers}")
return resnet(
units=units,
diff --git a/python/tvm/relay/testing/squeezenet.py
b/python/tvm/relay/testing/squeezenet.py
index 097f2230af..ce918fd879 100644
--- a/python/tvm/relay/testing/squeezenet.py
+++ b/python/tvm/relay/testing/squeezenet.py
@@ -32,10 +32,10 @@ from . import layers
# Helpers
def _make_fire(net, squeeze_channels, expand1x1_channels, expand3x3_channels,
prefix):
- net = _make_fire_conv(net, squeeze_channels, 1, 0, "%s_input" % prefix)
+ net = _make_fire_conv(net, squeeze_channels, 1, 0, f"{prefix}_input")
- left = _make_fire_conv(net, expand1x1_channels, 1, 0, "%s_left" % prefix)
- right = _make_fire_conv(net, expand3x3_channels, 3, 1, "%s_right" % prefix)
+ left = _make_fire_conv(net, expand1x1_channels, 1, 0, f"{prefix}_left")
+ right = _make_fire_conv(net, expand3x3_channels, 3, 1, f"{prefix}_right")
# NOTE : Assume NCHW layout here
net = relay.concatenate((left, right), axis=1)
return net
@@ -47,9 +47,9 @@ def _make_fire_conv(net, channels, kernel_size, padding=0,
prefix=""):
channels=channels,
kernel_size=(kernel_size, kernel_size),
padding=(padding, padding),
- name="%s_conv" % prefix,
+ name=f"{prefix}_conv",
)
- net = relay.nn.bias_add(net, relay.var("%s_conv_bias" % prefix))
+ net = relay.nn.bias_add(net, relay.var(f"{prefix}_conv_bias"))
net = relay.nn.relu(net)
return net
@@ -72,10 +72,9 @@ def get_net(batch_size, image_shape, num_classes, version,
dtype):
version : str, optional
"1.0" or "1.1" of SqueezeNet
"""
- assert version in [
- "1.0",
- "1.1",
- ], "Unsupported SqueezeNet version {version}:" "1.0 or 1.1
expected".format(version=version)
+ assert version in ["1.0", "1.1"], (
+ f"Unsupported SqueezeNet version {version}:" "1.0 or 1.1 expected"
+ )
data_shape = (batch_size,) + image_shape
net = relay.var("data", shape=data_shape, dtype=dtype)
if version == "1.0":
diff --git a/python/tvm/relay/testing/tf.py b/python/tvm/relay/testing/tf.py
index e09111a205..158de22eea 100644
--- a/python/tvm/relay/testing/tf.py
+++ b/python/tvm/relay/testing/tf.py
@@ -109,9 +109,9 @@ def vmobj_to_list(o):
elif "tensor" in o.constructor.name_hint:
result = [o.fields[0].numpy()]
else:
- raise RuntimeError("Unknown object type: %s" %
o.constructor.name_hint)
+ raise RuntimeError(f"Unknown object type:
{o.constructor.name_hint}")
else:
- raise RuntimeError("Unknown object type: %s" % type(o))
+ raise RuntimeError(f"Unknown object type: {type(o)}")
return result
@@ -134,9 +134,7 @@ def AddShapesToGraphDef(session, out_node):
"""
graph_def = tf_compat_v1.graph_util.convert_variables_to_constants(
- session,
- session.graph.as_graph_def(add_shapes=True),
- convert_to_list(out_node),
+ session, session.graph.as_graph_def(add_shapes=True),
convert_to_list(out_node)
)
return graph_def
diff --git a/python/tvm/relay/testing/tflite.py
b/python/tvm/relay/testing/tflite.py
index b698b004b4..df9c0bcadf 100644
--- a/python/tvm/relay/testing/tflite.py
+++ b/python/tvm/relay/testing/tflite.py
@@ -56,7 +56,7 @@ class TFLiteModel:
elif activation == "NONE":
pass
else:
- assert False, "Unsupported activation {}".format(activation)
+ assert False, f"Unsupported activation {activation}"
return op
return conv2d_single_function
diff --git a/python/tvm/relay/testing/vgg.py b/python/tvm/relay/testing/vgg.py
index b14c069ed0..426cd9e608 100644
--- a/python/tvm/relay/testing/vgg.py
+++ b/python/tvm/relay/testing/vgg.py
@@ -34,14 +34,14 @@ def get_feature(internal_layer, layers, filters,
batch_norm=False):
kernel_size=(3, 3),
padding=(1, 1),
channels=filters[i],
- name="conv%s_%s" % (i + 1, j + 1),
+ name=f"conv{i + 1}_{j + 1}",
)
internal_layer = relay.nn.bias_add(
- internal_layer, relay.var("conv%s_%s_bias" % (i + 1, j + 1))
+ internal_layer, relay.var(f"conv{i + 1}_{j + 1}_bias")
)
if batch_norm:
internal_layer = wrapper.batch_norm_infer(
- data=internal_layer, name="bn%s_%s" % (i + 1, j + 1)
+ data=internal_layer, name=f"bn{i + 1}_{j + 1}"
)
internal_layer = relay.nn.relu(data=internal_layer)
internal_layer = relay.nn.max_pool2d(data=internal_layer,
pool_size=(2, 2), strides=(2, 2))
@@ -90,7 +90,7 @@ def get_net(batch_size, image_shape, num_classes, dtype,
num_layers=11, batch_no
19: ([2, 2, 4, 4, 4], [64, 128, 256, 512, 512]),
}
if num_layers not in vgg_spec:
- raise ValueError("Invalide num_layers {}. Choices are
11,13,16,19.".format(num_layers))
+ raise ValueError(f"Invalid num_layers {num_layers}. Choices are
11,13,16,19.")
layers, filters = vgg_spec[num_layers]
data_shape = (batch_size,) + image_shape
data = relay.var("data", shape=data_shape, dtype=dtype)
diff --git a/python/tvm/relay/transform/fake_quantization_to_integer.py
b/python/tvm/relay/transform/fake_quantization_to_integer.py
index 7375a4f3c0..82255c5663 100644
--- a/python/tvm/relay/transform/fake_quantization_to_integer.py
+++ b/python/tvm/relay/transform/fake_quantization_to_integer.py
@@ -219,13 +219,7 @@ def bias_add(expr, type_map):
and tvm.ir.structural_equal(x_t.dtype, b_t.dtype)
):
b = relay.qnn.op.requantize(
- b,
- b_t.scale,
- b_t.zero_point,
- in_scale,
- in_zero_point,
- out_dtype=x_t.dtype,
- axis=0,
+ b, b_t.scale, b_t.zero_point, in_scale, in_zero_point,
out_dtype=x_t.dtype, axis=0
)
else:
# If the bias is a constant, we need to quantize it
@@ -522,15 +516,13 @@ def register_binary_qnn(op_name, op):
# addition is typically done in 32 bit).
return [left + right, left_t]
- assert (
- len(out_t.scale.data.shape) == 0
- ), "The output scale needs to be a scalar, but got a tensor of shape
{}".format(
- out_t.scale.data.shape
+ assert len(out_t.scale.data.shape) == 0, (
+ f"The output scale needs to be a scalar, but got a tensor of shape
"
+ f"{out_t.scale.data.shape}"
)
- assert (
- len(out_t.zero_point.data.shape) == 0
- ), "The output zero point needs to be a scalar, but got a tensor of
shape {}".format(
- out_t.zero_point.data.shape
+ assert len(out_t.zero_point.data.shape) == 0, (
+ f"The output zero point needs to be a scalar, but got a tensor of
shape "
+ f"{out_t.zero_point.data.shape}"
)
out = op(
@@ -601,13 +593,7 @@ def register_unary_qnn(op_name, op):
arg = expr.args[0]
x_t = type_map[arg]
out_t = type_map[expr]
- out = op(
- arg,
- x_t.scale,
- x_t.zero_point,
- out_t.scale,
- out_t.zero_point,
- )
+ out = op(arg, x_t.scale, x_t.zero_point, out_t.scale, out_t.zero_point)
return [out, out_t]
return register_fake_quantization_to_integer(op_name, unary)
diff --git a/python/tvm/relay/type_functor.py b/python/tvm/relay/type_functor.py
index 490464ba12..39f94aeca7 100644
--- a/python/tvm/relay/type_functor.py
+++ b/python/tvm/relay/type_functor.py
@@ -64,7 +64,7 @@ class TypeFunctor:
elif isinstance(typ, TypeData):
return self.visit_type_data(typ)
else:
- raise Exception("unhandled case: {0}".format(type(typ)))
+ raise Exception(f"unhandled case: {type(typ)}")
def visit_type_var(self, _):
raise NotImplementedError()
