[CI] Bump black version to 22.3.0 (#10960)

* Make all required adjusts in the code to comply with the new version
* Upadte ci-lint to v0.71, based on tlcpackstaging/ci_lint:20220411-060305-45f3d4a52

Jenkinsfile

@@ -45,11 +45,11 @@
 // 'python3 jenkins/generate.py'
 // Note: This timestamp is here to ensure that updates to the Jenkinsfile are
 // always rebased on main before merging:
-// Generated at 2022-04-07T13:50:22.427152
+// Generated at 2022-04-11T10:45:26.226802
 
 import org.jenkinsci.plugins.pipeline.modeldefinition.Utils
 // NOTE: these lines are scanned by docker/dev_common.sh. Please update the regex as needed. -->
-ci_lint = 'tlcpack/ci-lint:v0.69'
+ci_lint = 'tlcpack/ci-lint:v0.71'
 ci_gpu = 'tlcpack/ci-gpu:v0.84'
 ci_cpu = 'tlcpack/ci-cpu:v0.83'
 ci_wasm = 'tlcpack/ci-wasm:v0.73'

apps/topi_recipe/gemm/android_gemm_square.py

@@ -34,7 +34,7 @@
 
 
 def ngflops(N):
-    return 2.0 * float(N * N * N) / (10 ** 9)
+    return 2.0 * float(N * N * N) / (10**9)
 
 
 dtype = "float32"

jenkins/Jenkinsfile.j2

@@ -51,7 +51,7 @@ import org.jenkinsci.plugins.pipeline.modeldefinition.Utils
 {% import 'jenkins/macros.j2' as m with context -%}
 
 // NOTE: these lines are scanned by docker/dev_common.sh. Please update the regex as needed. -->
-ci_lint = 'tlcpack/ci-lint:v0.69'
+ci_lint = 'tlcpack/ci-lint:v0.71'
 ci_gpu = 'tlcpack/ci-gpu:v0.84'
 ci_cpu = 'tlcpack/ci-cpu:v0.83'
 ci_wasm = 'tlcpack/ci-wasm:v0.73'

python/tvm/autotvm/task/space.py

@@ -187,7 +187,7 @@ def get_pow2s(n):
     factors: list
         List of all power-of-two numbers
     """
-    return [2 ** x for x in range(math.floor(math.log2(n)) + 1)]
+    return [2**x for x in range(math.floor(math.log2(n)) + 1)]
 
 
 class SplitSpace(TransformSpace):

python/tvm/contrib/debugger/debug_result.py

@@ -154,7 +154,7 @@ def dump_chrome_trace(self):
         """Dump the trace to the Chrome trace.json format."""
 
         def s_to_us(t):
-            return t * 10 ** 6
+            return t * 10**6
 
         starting_times = np.zeros(len(self._time_list) + 1)
         starting_times[1:] = np.cumsum([times[0] for times in self._time_list])

python/tvm/relay/frontend/paddlepaddle.py

@@ -658,7 +658,7 @@ def convert_gelu(g, op, block):
     x = g.get_node(op.input("X")[0])
     out = x * (
         _expr.const(0.5, dtype="float32")
-        + _op.erf(x * _expr.const(0.5 ** 0.5, dtype="float32")) * _expr.const(0.5, dtype="float32")
+        + _op.erf(x * _expr.const(0.5**0.5, dtype="float32")) * _expr.const(0.5, dtype="float32")
     )
     g.add_node(op.output("Out")[0], out)
 

python/tvm/relay/frontend/pytorch.py

@@ -827,7 +827,7 @@ def gelu(self, inputs, input_types):
         # with tanh and third order polynomials, but this is "true" gelu
         return data * (
             _expr.const(0.5, dtype=dtype)
-            + _op.erf(data * _expr.const(0.5 ** 0.5, dtype=dtype)) * _expr.const(0.5, dtype=dtype)
+            + _op.erf(data * _expr.const(0.5**0.5, dtype=dtype)) * _expr.const(0.5, dtype=dtype)
         )
 
     def selu(self, inputs, input_types):

python/tvm/relay/qnn/op/canonicalizations.py

@@ -75,7 +75,7 @@ def create_integer_lookup_table(
     # inputs_quantized = np.array(range(dtype_info.min, dtype_info.max + 1)).astype(in_dtype)
 
     # First generate a list of all num_bit integer patterns
-    inputs_quantized = np.array(range(0, 2 ** num_bits), dtype=f"uint{num_bits}")
+    inputs_quantized = np.array(range(0, 2**num_bits), dtype=f"uint{num_bits}")
 
     # Reinterpret bits as the real datatype
     # Note what we are doing here is a bit tricky, the canonical view of our lookup table

python/tvm/relay/quantize/_calibrate.py

@@ -159,7 +159,7 @@ def visit_func(expr):
             def _make_const(val):
                 return _expr.const(val, "float32")
 
-            valid_range = 2 ** valid_bit
+            valid_range = 2**valid_bit
             const_params[ndom_scale] = _make_const(scale / valid_range)
             const_params[nclip_min] = _make_const(-(valid_range - 1))
             const_params[nclip_max] = _make_const((valid_range - 1))

python/tvm/relay/testing/tf.py

@@ -321,7 +321,7 @@ def pick_from_weight(weight, pows=1.0):
     """Identify token from Softmax output.
     This token will be mapped to word in the vocabulary.
     """
-    weight = weight ** pows
+    weight = weight**pows
     t = np.cumsum(weight)
     s = np.sum(weight)
     return int(np.searchsorted(t, 0.5 * s))

python/tvm/testing/utils.py

@@ -218,7 +218,7 @@ def compare_derivative(j, n_der, grad):
         wrong_percentage = int(100 * len(wrong_positions) / np.prod(grad.shape))
 
         dist = np.sqrt(np.sum((ngrad - grad) ** 2))
-        grad_norm = np.sqrt(np.sum(ngrad ** 2))
+        grad_norm = np.sqrt(np.sum(ngrad**2))
 
         if not (np.isfinite(dist) and np.isfinite(grad_norm)):
             raise ValueError(

python/tvm/tir/schedule/_type_checker.py

@@ -57,7 +57,6 @@ def union(type_: Any) -> Optional[List[type]]:
                     return list(subtypes)
             return None
 
-
 elif hasattr(typing, "_Union"):
 
     class _Subtype:  # type: ignore

python/tvm/topi/gpu/dense.py

@@ -153,8 +153,8 @@ def _schedule_dense_large_batch(cfg, s, C):
     # create tuning space
     try:
         block_cand = [64, 128]
-        vthread_cand = [2 ** x for x in range(1, 7)]
-        n_thread_cand = [2 ** x for x in range(3, 7)]
+        vthread_cand = [2**x for x in range(1, 7)]
+        n_thread_cand = [2**x for x in range(3, 7)]
         cfg.define_split(
             "tile_x",
             batch,

python/tvm/topi/random/kernel.py

@@ -233,7 +233,7 @@ def threefry_generate(gen, out_shape):
     for s in out_shape:
         out_len *= s
     assert (
-        out_len.value <= 2 ** 64 - 1
+        out_len.value <= 2**64 - 1
     ), f"Can only generate up to 2^64 random numbers, but {out_len} were requested."
 
     def gen_ir(gen_ptr, out_gen_ptr, out_array_ptr):
@@ -264,7 +264,7 @@ def gen_ir(gen_ptr, out_gen_ptr, out_array_ptr):
 
         # Max value for counter should be 2**64-2 because we need to reserve a special value to
         # indicate the counter is used up.
-        with irb.if_scope(gen[7] < tir.const(2 ** 64 - 1, dtype=gen.dtype) - out_len):
+        with irb.if_scope(gen[7] < tir.const(2**64 - 1, dtype=gen.dtype) - out_len):
             for i in range(10):
                 tmp[i] = gen[i]
         with irb.else_scope():

python/tvm/topi/testing/correlation_nchw_python.py

@@ -103,5 +103,5 @@ def correlation_nchw_python(
                                     pad_data2[nbatch, channel, y2 + h, x2 + w],
                                 )
 
-    out /= float(kernel_size ** 2 * data1.shape[1])
+    out /= float(kernel_size**2 * data1.shape[1])
     return out

tests/python/contrib/test_cmsisnn/utils.py

@@ -290,7 +290,7 @@ def generate_ref_data_tflite(model):
 
 
 def create_conv2d_tflite_model(ifm_shape, kernel_shape, strides, dilation, padding, activation):
-    """ This method prepares TFlite graph with a single Conv2d layer """
+    """This method prepares TFlite graph with a single Conv2d layer"""
     import tensorflow as tf
 
     class Model(tf.Module):

tests/python/contrib/test_ethosu/cascader/conftest.py

@@ -29,7 +29,7 @@
 def FLASH():
     return cs.MemoryRegion(
         name="FLASH",
-        size=10 ** 7,
+        size=10**7,
         read_bandwidth=4,
         write_bandwidth=4,
         read_latency=0,
@@ -42,7 +42,7 @@ def FLASH():
 def DRAM():
     return cs.MemoryRegion(
         name="DRAM",
-        size=10 ** 9,
+        size=10**9,
         read_bandwidth=8,
         write_bandwidth=8,
         read_latency=0,
@@ -55,7 +55,7 @@ def DRAM():
 def SRAM():
     return cs.MemoryRegion(
         name="SRAM",
-        size=10 ** 6,
+        size=10**6,
         read_bandwidth=16,
         write_bandwidth=16,
         read_latency=0,

tests/python/frontend/pytorch/test_forward.py

@@ -2193,7 +2193,7 @@ def test_vgg11_bn():
 def test_custom_conversion_map():
     def get_roi_align():
         pool_size = 5
-        n_channels = 2 * (pool_size ** 2)
+        n_channels = 2 * (pool_size**2)
         x = torch.rand(2, n_channels, 10, 10)
         rois = torch.tensor(
             [

tests/python/relay/test_op_grad_level1.py

@@ -56,11 +56,11 @@ class TestUnaryOp:
         "log10": (tvm.relay.log10, lambda x, g: g * (1 / (np.log(10) * x))),
         "cosh": (tvm.relay.cosh, lambda x, g: g * (np.sinh(x))),
         "sinh": (tvm.relay.sinh, lambda x, g: g * (np.cosh(x))),
-        "asin": (tvm.relay.asin, lambda x, g: g * (1.0 / (1.0 - x ** 2) ** (1.0 / 2.0))),
-        "acos": (tvm.relay.acos, lambda x, g: g * (-1.0 / (1.0 - x ** 2.0) ** (1.0 / 2.0))),
-        "acosh": (tvm.relay.acosh, lambda x, g: g * (1.0 / (x ** 2 - 1.0) ** (1.0 / 2.0))),
-        "asinh": (tvm.relay.asinh, lambda x, g: g * (1.0 / (x ** 2 + 1.0) ** (1.0 / 2.0))),
-        "atanh": (tvm.relay.atanh, lambda x, g: g * (-1.0 / (x ** 2 - 1.0))),
+        "asin": (tvm.relay.asin, lambda x, g: g * (1.0 / (1.0 - x**2) ** (1.0 / 2.0))),
+        "acos": (tvm.relay.acos, lambda x, g: g * (-1.0 / (1.0 - x**2.0) ** (1.0 / 2.0))),
+        "acosh": (tvm.relay.acosh, lambda x, g: g * (1.0 / (x**2 - 1.0) ** (1.0 / 2.0))),
+        "asinh": (tvm.relay.asinh, lambda x, g: g * (1.0 / (x**2 + 1.0) ** (1.0 / 2.0))),
+        "atanh": (tvm.relay.atanh, lambda x, g: g * (-1.0 / (x**2 - 1.0))),
     }
 
     relay_op, ref_func = tvm.testing.parameters(*config.values(), ids=config.keys())
@@ -136,7 +136,7 @@ class TestBinaryOp:
         "add": (relay.add, lambda x, y: [np.ones_like(x), np.ones_like(y)]),
         "subtract": (relay.subtract, lambda x, y: [np.ones_like(x), -np.ones_like(y)]),
         "multiply": (relay.multiply, lambda x, y: [y, x]),
-        "divide": (relay.divide, lambda x, y: [1 / y, -x / (y ** 2)]),
+        "divide": (relay.divide, lambda x, y: [1 / y, -x / (y**2)]),
     }
 
     relay_op, ref_func = tvm.testing.parameters(*config.values(), ids=config.keys())

tests/python/topi/python/test_topi_prng.py

@@ -120,14 +120,14 @@ def test_threefry_generate(target, dev):
 
     # test enough generates to go over generate limit
     gen = np.array(
-        [0, 0, 0, 0, 0, 0, 0, 2 ** 64 - 2, 1 << 63, 0], dtype="uint64"
+        [0, 0, 0, 0, 0, 0, 0, 2**64 - 2, 1 << 63, 0], dtype="uint64"
     )  # make counter large
     a, rands = threefry_generate(target, dev, gen, (2048,))
     assert gen[4] != a[4], "Overflow of counter should trigger path change"
     assert a[7] == 2048, "Overflow of counter should still update counter"
 
     # check generate with path at length limit
-    gen = np.array([0, 0, 0, 0, 0, 0, 0, 2 ** 64 - 2, 0, 0], dtype="uint64")  # make counter large
+    gen = np.array([0, 0, 0, 0, 0, 0, 0, 2**64 - 2, 0, 0], dtype="uint64")  # make counter large
     a, rands = threefry_generate(target, dev, gen, (2048,))
     assert (
         gen[0:4] != a[0:4]

tests/python/topi/python/test_topi_transform.py

@@ -861,10 +861,10 @@ def test_reinterpret():
         (1000,), "int16", "uint16", lambda shape: np.random.randint(-1000, 1000, size=shape)
     )
     verify_reinterpret(
-        (1000,), "uint32", "int32", lambda shape: np.random.randint(0, 2 ** 32 - 1, size=shape)
+        (1000,), "uint32", "int32", lambda shape: np.random.randint(0, 2**32 - 1, size=shape)
     )
     verify_reinterpret(
-        (1000,), "uint32", "int32", lambda shape: np.random.randint(0, 2 ** 32 - 1, size=shape)
+        (1000,), "uint32", "int32", lambda shape: np.random.randint(0, 2**32 - 1, size=shape)
     )
 
 

tests/python/unittest/test_arith_canonical_simplify.py

@@ -331,7 +331,7 @@ def test_simplify_cast():
     # cast(i32, i + j - 100)
     i = te.var("i", dtype="int64")
     j = te.var("j", dtype="int64")
-    ck.analyzer.update(i, tvm.arith.ConstIntBound(0, 2 ** 31 - 1))
+    ck.analyzer.update(i, tvm.arith.ConstIntBound(0, 2**31 - 1))
     ck.analyzer.update(j, tvm.arith.ConstIntBound(0, 10))
     res = tcast("int32", i + j - 100)
     ck.verify(res, res)

tests/python/unittest/test_auto_scheduler_compute_dag.py

@@ -47,25 +47,25 @@ def test_estimate_flop():
     N = 512
     A, B, C = matmul_auto_scheduler_test(N, N, N)
     dag = auto_scheduler.ComputeDAG([A, B, C])
-    assert abs(dag.flop_ct - 2 * N ** 3) < 0.5
+    assert abs(dag.flop_ct - 2 * N**3) < 0.5
 
     D = topi.nn.relu(C)
     dag = auto_scheduler.ComputeDAG([A, B, D])
-    assert abs(dag.flop_ct - (2 * N ** 3 + N * N)) < 0.5
+    assert abs(dag.flop_ct - (2 * N**3 + N * N)) < 0.5
 
     # should not count the comparison operations in padding
     E = topi.nn.pad(C, [1, 1])
     dag = auto_scheduler.ComputeDAG([A, B, E])
-    assert abs(dag.flop_ct - 2 * N ** 3) < 0.5
+    assert abs(dag.flop_ct - 2 * N**3) < 0.5
 
     F = te.compute((N, N), lambda i, j: E[i, j], name="F", attrs={"FLOP": 1234})
     dag = auto_scheduler.ComputeDAG([A, B, F])
-    assert abs(dag.flop_ct - (2 * N ** 3 + 1234)) < 0.5
+    assert abs(dag.flop_ct - (2 * N**3 + 1234)) < 0.5
 
     A = te.placeholder((N, N), dtype="float32", name="A")
     F = te.compute((N, N), lambda i, j: te.if_then_else(A[i, j] > 0, A[i, j], 0))
     dag = auto_scheduler.ComputeDAG([A, F])
-    assert abs(dag.flop_ct - N ** 2) < 0.5
+    assert abs(dag.flop_ct - N**2) < 0.5
 
 
 def test_stage_order():

tests/python/unittest/test_auto_scheduler_feature.py

@@ -78,8 +78,8 @@ def test_cpu_matmul():
     """
 
     # check touched memory in bytes, touched unique memory in bytes, reuse distance, etc.
-    assert fequal(fea_dict[c_name + ".bytes"], math.log2(512 ** 3 * 4 + 1))
-    assert fequal(fea_dict[b_name + ".unique_bytes"], math.log2(512 ** 2 * 4 + 1))
+    assert fequal(fea_dict[c_name + ".bytes"], math.log2(512**3 * 4 + 1))
+    assert fequal(fea_dict[b_name + ".unique_bytes"], math.log2(512**2 * 4 + 1))
     assert fequal(fea_dict[c_name + ".reuse_dis_iter"], math.log2(8 * 16 + 1))
     assert fequal(fea_dict[c_name + ".reuse_dis_bytes"], math.log2((8 * 16 + 8 + 16) * 4 + 1))
     assert fequal(fea_dict[c_name + ".reuse_ct"], math.log2(512 + 1))

tests/python/unittest/test_autotvm_space.py

@@ -76,7 +76,7 @@ def count4(n):
     # test overflow
     n = 25
     cfg = ConfigSpace()
-    cfg.define_split("x", cfg.axis(2 ** n), policy="factors", num_outputs=4)
+    cfg.define_split("x", cfg.axis(2**n), policy="factors", num_outputs=4)
     # count4(25) is 3276.
     assert len(cfg.space_map["x"]) == count4(n)
 

tests/python/unittest/test_format_si_prefix.py

@@ -30,7 +30,7 @@ def test_format_si_prefix():
     for i, prefix in enumerate(SI_PREFIXES):
         integer, decimal = random.randint(0, 1000), random.randint(0, 1000)
         exp = -24 + 3 * i  # 0th prefix (yocto) is 10^-24
-        number = integer * (10 ** exp) + decimal * (10 ** (exp - 3))
+        number = integer * (10**exp) + decimal * (10 ** (exp - 3))
         expected = integer + decimal / 1000
         assert isclose(utils.format_si_prefix(number, prefix), expected)
 

tests/python/unittest/test_target_codegen_c_host.py

@@ -111,7 +111,7 @@ def check_c():
         fadd = m["test_reinterpret"]
         dev = tvm.cpu(0)
         n = nn
-        a = tvm.nd.array(np.random.randint(-(2 ** 30), 2 ** 30, size=n).astype(A.dtype), dev)
+        a = tvm.nd.array(np.random.randint(-(2**30), 2**30, size=n).astype(A.dtype), dev)
         b = tvm.nd.array(np.zeros(n, dtype=B.dtype), dev)
         fadd(a, b)
         tvm.testing.assert_allclose(b.numpy(), (2 + a.numpy()).view("float32"))

tests/python/unittest/test_target_codegen_rocm.py

@@ -105,7 +105,7 @@ def check_rocm(dtype, n):
         dtype = np.random.choice(["float32", "float16", "int8", "int32"])
         logN = np.random.randint(1, 15)
         peturb = np.random.uniform(low=0.5, high=1.5)
-        check_rocm(dtype, int(peturb * (2 ** logN)))
+        check_rocm(dtype, int(peturb * (2**logN)))
 
 
 @tvm.testing.requires_rocm