def run(
N,
IC,
OC,
IH,
IW,
KH,
KW,
PH,
PW,
SH,
SW,
has_bias=True,
):
inp_v = np.random.normal(size=(N, IC, IH, IW))
w_v = np.random.normal(size=(N, OC, IC, KH, KW))
b_v = np.random.normal(size=(1, OC, 1, 1))
inp_scale = dtype.get_scale(inp_dtype)
w_scale = dtype.get_scale(w_dtype)
b_scale = dtype.get_scale(b_dtype)
inpv = dtype.convert_to_qint8(inp_v * inp_scale, inp_dtype)
wv = dtype.convert_to_qint8(w_v * w_scale, w_dtype)
bv = dtype.convert_to_qint32(b_v * b_scale, b_dtype)
inp_int8 = tensor(inpv, dtype=inp_dtype)
w_int8 = Parameter(wv, dtype=w_dtype)
b_int32 = Parameter(bv, dtype=b_dtype)
inp_fp32 = inp_int8.astype("float32")
w_fp32 = w_int8.astype("float32")
b_fp32 = b_int32.astype("float32")
def run_batch_conv_bias(inp, w, b):
b = b if has_bias else Parameter(np.zeros_like(b.numpy()))
result = F.quantized.batch_conv_bias_activation(
inp,
w,
b,
stride=(SH, SW),
padding=(PH, PW),
dtype=out_dtype,
)
return result.astype("float32")
expected = F.conv2d(inp_fp32, w_fp32[0],
b_fp32 if has_bias else None)[0]
expected = expected.astype(out_dtype).astype("float32")
expected = F.flatten(expected)
result = run_batch_conv_bias(inp_int8, w_int8, b_int32)
result = F.flatten(result)
np.testing.assert_allclose(result.numpy(),
expected.numpy(),
atol=outp_scale)
def test_dtype_int8_ffi_handle():
device = "xpux"
shape = (3, 3, 3)
data = np.random.random(shape).astype(np.float32) * 5 - 1
def identity(x):
return x
dtype = quint8(0.01, 127)
inp = convert_to_quint8(data, dtype)
oup = _get_compiled_result(inp, dtype, shape, device, calc_func=identity)
_check_result_attr(oup, dtype, "quint8")
np.testing.assert_allclose(convert_from_quint8(oup), convert_from_quint8(inp))
dtype = qint8(0.01)
inp = convert_to_qint8(data, dtype)
oup = _get_compiled_result(inp, dtype, shape, device, calc_func=identity)
_check_result_attr(oup, dtype, "qint8", is_unsigned=False)
np.testing.assert_allclose(convert_from_qint8(oup), convert_from_qint8(inp))
def run(
N,
IC,
OC,
IH,
IW,
KH,
KW,
PH,
PW,
SH,
SW,
has_bias=True,
nonlinear_mode="identity",
):
inp_v = np.random.normal(size=(N, IC, IH, IW))
w_v = np.random.normal(size=(OC, IC, KH, KW))
b_v = np.random.normal(size=(1, OC, 1, 1))
inp_scale = dtype.get_scale(inp_dtype)
w_scale = dtype.get_scale(w_dtype)
b_scale = dtype.get_scale(b_dtype)
inpv = dtype.convert_to_qint8(inp_v * inp_scale, inp_dtype)
wv = dtype.convert_to_qint8(w_v * w_scale, w_dtype)
bv = dtype.convert_to_qint32(b_v * b_scale, b_dtype)
inp_int8 = tensor(inpv, dtype=inp_dtype)
w_int8 = Parameter(wv, dtype=w_dtype)
b_int32 = Parameter(bv, dtype=b_dtype)
inp_fp32 = inp_int8.astype("float32")
w_fp32 = w_int8.astype("float32")
b_fp32 = b_int32.astype("float32")
def convert_to_nchw4(var):
var = F.reshape(var, (var.shape[0], var.shape[1] // 4, 4,
var.shape[2], var.shape[3]))
var = F.transpose(var, (0, 1, 3, 4, 2))
return var
def run_conv2d(inp, w, b):
O = F.conv2d(
inp,
w,
b if has_bias else None,
stride=(SH, SW),
padding=(PH, PW),
)
if nonlinear_mode == "relu":
return F.relu(O)
else:
return O
def run_conv_bias(inp, w, b, format="NCHW"):
b = b if has_bias else Parameter(np.zeros_like(b.numpy()))
if format == "NCHW4":
inp = convert_to_nchw4(inp)
w = convert_to_nchw4(w)
b = convert_to_nchw4(b)
return F.quantized.conv_bias_activation(
inp,
w,
b,
stride=(SH, SW),
padding=(PH, PW),
dtype=out_dtype,
nonlinear_mode=nonlinear_mode,
)
format = "NCHW4" if is_cuda_available() else "NCHW"
expected = run_conv2d(inp_fp32, w_fp32, b_fp32)
expected = expected.astype(out_dtype).astype("float32")
result = run_conv_bias(inp_int8, w_int8, b_int32,
format=format).astype("float32")
if format == "NCHW4":
result = F.transpose(result, (0, 1, 4, 2, 3))
expected = F.flatten(expected)
result = F.flatten(result)
np.testing.assert_allclose(result.numpy(),
expected.numpy(),
atol=outp_scale)
def test_func(
N,
IC,
IH,
IW,
OC,
KH,
KW,
SH,
SW,
PH,
PW,
DH,
DW,
groups=1,
has_bias=True,
conv_mode: str = "cross_correlation",
compute_mode: str = "default",
):
inp_scale = np.float32(rng.uniform(low=0.04, high=0.06))
weight_scale = np.float32(rng.uniform(low=0.04, high=0.06))
bias_scale = inp_scale * weight_scale
out_scale = np.float32(rng.uniform(low=0.04, high=0.06))
inp_dtype = dtype.qint8(inp_scale)
weight_dtype = dtype.qint8(weight_scale)
bias_dtype = dtype.qint32(bias_scale)
out_dtype = dtype.qint8(out_scale)
inp_fp32 = rng.uniform(low=-1, high=1,
size=(N, IC, IH, IW)).astype(np.float32)
weight_fp32 = rng.uniform(low=-1, high=1,
size=(IC, OC, KH, KW)).astype(np.float32)
bias_fp32 = rng.uniform(low=-1, high=1,
size=(1, OC, 1, 1)).astype(np.float32)
inp_int8 = dtype.convert_to_qint8(inp_fp32, inp_dtype)
weight_int8 = dtype.convert_to_qint8(weight_fp32, weight_dtype)
bias_int32 = dtype.convert_to_qint32(bias_fp32, bias_dtype)
inp_int8 = mge.tensor(inp_int8, dtype=inp_dtype)
weight_int8 = mge.Parameter(weight_int8, dtype=weight_dtype)
bias_int32 = mge.Parameter(bias_int32, dtype=bias_dtype)
inp_fp32 = inp_int8.astype("float32")
weight_fp32 = weight_int8.astype("float32")
bias_fp32 = bias_int32.astype("float32")
expected = F.conv_transpose2d(
inp_fp32,
weight_fp32,
bias_fp32 if has_bias else None,
stride=(SH, SW),
padding=(PH, PW),
dilation=(DH, DW),
groups=groups,
conv_mode=conv_mode,
compute_mode=compute_mode,
)
expected = dtype.convert_to_qint8(expected.numpy(), out_dtype)
expected = dtype.convert_from_qint8(expected)
conv_transpose2d = ConvTranspose2d(
in_channels=IC,
out_channels=OC,
kernel_size=(KH, KW),
stride=(SH, SW),
padding=(PH, PW),
dilation=(DH, DW),
groups=groups,
bias=has_bias,
conv_mode=conv_mode,
compute_mode=compute_mode,
dtype=out_dtype,
)
conv_transpose2d.weight = mge.Parameter(weight_int8)
if has_bias:
conv_transpose2d.bias = mge.Parameter(bias_int32)
result = conv_transpose2d.forward(inp_int8).numpy()
result = dtype.convert_from_qint8(result)
np.testing.assert_allclose(result, expected, atol=out_scale)
