mirror of
https://github.com/pytorch/pytorch.git
synced 2025-10-20 21:14:14 +08:00
[codemod][lint][fbcode/c*] Enable BLACK by default
Test Plan: manual inspection & sandcastle Reviewed By: zertosh Differential Revision: D30279364 fbshipit-source-id: c1ed77dfe43a3bde358f92737cd5535ae5d13c9a
This commit is contained in:
Zsolt Dollenstein
committed by
Facebook GitHub Bot
Facebook GitHub Bot
parent
aac3c7bd06
commit
b004307252
@ -1,11 +1,12 @@
|
||||
import io
|
||||
import numpy as np
|
||||
import os
|
||||
import shutil
|
||||
import sys
|
||||
import unittest
|
||||
import uuid
|
||||
|
||||
import numpy as np
|
||||
|
||||
TEST_TENSORBOARD = True
|
||||
try:
|
||||
import tensorboard.summary.writer.event_file_writer # noqa: F401
|
||||
@ -31,8 +32,9 @@ skipIfNoCaffe2 = unittest.skipIf(not TEST_CAFFE2, "no caffe2")
|
||||
TEST_MATPLOTLIB = True
|
||||
try:
|
||||
import matplotlib
|
||||
if os.environ.get('DISPLAY', '') == '':
|
||||
matplotlib.use('Agg')
|
||||
|
||||
if os.environ.get("DISPLAY", "") == "":
|
||||
matplotlib.use("Agg")
|
||||
import matplotlib.pyplot as plt
|
||||
except ImportError:
|
||||
TEST_MATPLOTLIB = False
|
||||
@ -41,13 +43,16 @@ skipIfNoMatplotlib = unittest.skipIf(not TEST_MATPLOTLIB, "no matplotlib")
|
||||
import torch
|
||||
from torch.testing._internal.common_utils import TestCase, run_tests, TEST_WITH_ASAN
|
||||
|
||||
|
||||
def tensor_N(shape, dtype=float):
|
||||
numel = np.prod(shape)
|
||||
x = (np.arange(numel, dtype=dtype)).reshape(shape)
|
||||
return x
|
||||
|
||||
|
||||
class BaseTestCase(TestCase):
|
||||
""" Base class used for all TensorBoard tests """
|
||||
"""Base class used for all TensorBoard tests"""
|
||||
|
||||
def setUp(self):
|
||||
if not TEST_TENSORBOARD:
|
||||
return self.skipTest("Skip the test since TensorBoard is not installed")
|
||||
@ -67,14 +72,15 @@ class BaseTestCase(TestCase):
|
||||
|
||||
|
||||
if TEST_TENSORBOARD:
|
||||
from tensorboard.compat.proto.graph_pb2 import GraphDef
|
||||
from torch.utils.tensorboard import summary, SummaryWriter
|
||||
from torch.utils.tensorboard._utils import _prepare_video, convert_to_HWC
|
||||
from torch.utils.tensorboard._convert_np import make_np
|
||||
from torch.utils.tensorboard import _caffe2_graph as c2_graph
|
||||
from torch.utils.tensorboard._pytorch_graph import graph
|
||||
from google.protobuf import text_format
|
||||
from PIL import Image
|
||||
from tensorboard.compat.proto.graph_pb2 import GraphDef
|
||||
from torch.utils.tensorboard import _caffe2_graph as c2_graph
|
||||
from torch.utils.tensorboard import summary, SummaryWriter
|
||||
from torch.utils.tensorboard._convert_np import make_np
|
||||
from torch.utils.tensorboard._pytorch_graph import graph
|
||||
from torch.utils.tensorboard._utils import _prepare_video, convert_to_HWC
|
||||
|
||||
|
||||
class TestTensorBoardPyTorchNumpy(BaseTestCase):
|
||||
def test_pytorch_np(self):
|
||||
@ -92,7 +98,9 @@ class TestTensorBoardPyTorchNumpy(BaseTestCase):
|
||||
|
||||
# CUDA variable
|
||||
if torch.cuda.device_count() > 0:
|
||||
self.assertIsInstance(make_np(torch.autograd.Variable(tensor).cuda()), np.ndarray)
|
||||
self.assertIsInstance(
|
||||
make_np(torch.autograd.Variable(tensor).cuda()), np.ndarray
|
||||
)
|
||||
|
||||
# python primitive type
|
||||
self.assertIsInstance(make_np(0), np.ndarray)
|
||||
@ -104,12 +112,12 @@ class TestTensorBoardPyTorchNumpy(BaseTestCase):
|
||||
|
||||
def test_pytorch_write(self):
|
||||
with self.createSummaryWriter() as w:
|
||||
w.add_scalar('scalar', torch.autograd.Variable(torch.rand(1)), 0)
|
||||
w.add_scalar("scalar", torch.autograd.Variable(torch.rand(1)), 0)
|
||||
|
||||
def test_pytorch_histogram(self):
|
||||
with self.createSummaryWriter() as w:
|
||||
w.add_histogram('float histogram', torch.rand((50,)))
|
||||
w.add_histogram('int histogram', torch.randint(0, 100, (50,)))
|
||||
w.add_histogram("float histogram", torch.rand((50,)))
|
||||
w.add_histogram("int histogram", torch.randint(0, 100, (50,)))
|
||||
|
||||
def test_pytorch_histogram_raw(self):
|
||||
with self.createSummaryWriter() as w:
|
||||
@ -118,52 +126,59 @@ class TestTensorBoardPyTorchNumpy(BaseTestCase):
|
||||
bins = [0.0, 0.25, 0.5, 0.75, 1.0]
|
||||
counts, limits = np.histogram(floats, bins)
|
||||
sum_sq = floats.dot(floats).item()
|
||||
w.add_histogram_raw('float histogram raw',
|
||||
min=floats.min().item(),
|
||||
max=floats.max().item(),
|
||||
num=num,
|
||||
sum=floats.sum().item(),
|
||||
sum_squares=sum_sq,
|
||||
bucket_limits=limits[1:].tolist(),
|
||||
bucket_counts=counts.tolist())
|
||||
w.add_histogram_raw(
|
||||
"float histogram raw",
|
||||
min=floats.min().item(),
|
||||
max=floats.max().item(),
|
||||
num=num,
|
||||
sum=floats.sum().item(),
|
||||
sum_squares=sum_sq,
|
||||
bucket_limits=limits[1:].tolist(),
|
||||
bucket_counts=counts.tolist(),
|
||||
)
|
||||
|
||||
ints = make_np(torch.randint(0, 100, (num,)))
|
||||
bins = [0, 25, 50, 75, 100]
|
||||
counts, limits = np.histogram(ints, bins)
|
||||
sum_sq = ints.dot(ints).item()
|
||||
w.add_histogram_raw('int histogram raw',
|
||||
min=ints.min().item(),
|
||||
max=ints.max().item(),
|
||||
num=num,
|
||||
sum=ints.sum().item(),
|
||||
sum_squares=sum_sq,
|
||||
bucket_limits=limits[1:].tolist(),
|
||||
bucket_counts=counts.tolist())
|
||||
w.add_histogram_raw(
|
||||
"int histogram raw",
|
||||
min=ints.min().item(),
|
||||
max=ints.max().item(),
|
||||
num=num,
|
||||
sum=ints.sum().item(),
|
||||
sum_squares=sum_sq,
|
||||
bucket_limits=limits[1:].tolist(),
|
||||
bucket_counts=counts.tolist(),
|
||||
)
|
||||
|
||||
ints = torch.tensor(range(0, 100)).float()
|
||||
nbins = 100
|
||||
counts = torch.histc(ints, bins=nbins, min=0, max=99)
|
||||
limits = torch.tensor(range(nbins))
|
||||
sum_sq = ints.dot(ints).item()
|
||||
w.add_histogram_raw('int histogram raw',
|
||||
min=ints.min().item(),
|
||||
max=ints.max().item(),
|
||||
num=num,
|
||||
sum=ints.sum().item(),
|
||||
sum_squares=sum_sq,
|
||||
bucket_limits=limits.tolist(),
|
||||
bucket_counts=counts.tolist())
|
||||
w.add_histogram_raw(
|
||||
"int histogram raw",
|
||||
min=ints.min().item(),
|
||||
max=ints.max().item(),
|
||||
num=num,
|
||||
sum=ints.sum().item(),
|
||||
sum_squares=sum_sq,
|
||||
bucket_limits=limits.tolist(),
|
||||
bucket_counts=counts.tolist(),
|
||||
)
|
||||
|
||||
|
||||
class TestTensorBoardUtils(BaseTestCase):
|
||||
def test_to_HWC(self):
|
||||
test_image = np.random.randint(0, 256, size=(3, 32, 32), dtype=np.uint8)
|
||||
converted = convert_to_HWC(test_image, 'chw')
|
||||
converted = convert_to_HWC(test_image, "chw")
|
||||
self.assertEqual(converted.shape, (32, 32, 3))
|
||||
test_image = np.random.randint(0, 256, size=(16, 3, 32, 32), dtype=np.uint8)
|
||||
converted = convert_to_HWC(test_image, 'nchw')
|
||||
converted = convert_to_HWC(test_image, "nchw")
|
||||
self.assertEqual(converted.shape, (64, 256, 3))
|
||||
test_image = np.random.randint(0, 256, size=(32, 32), dtype=np.uint8)
|
||||
converted = convert_to_HWC(test_image, 'hw')
|
||||
converted = convert_to_HWC(test_image, "hw")
|
||||
self.assertEqual(converted.shape, (32, 32, 3))
|
||||
|
||||
def test_convert_to_HWC_dtype_remains_same(self):
|
||||
@ -171,10 +186,13 @@ class TestTensorBoardUtils(BaseTestCase):
|
||||
# thus the scale_factor calculated for the image is 1
|
||||
test_image = torch.tensor([[[[1, 2, 3], [4, 5, 6]]]], dtype=torch.uint8)
|
||||
tensor = make_np(test_image)
|
||||
tensor = convert_to_HWC(tensor, 'NCHW')
|
||||
tensor = convert_to_HWC(tensor, "NCHW")
|
||||
scale_factor = summary._calc_scale_factor(tensor)
|
||||
self.assertEqual(scale_factor, 1, msg='Values are already in [0, 255], scale factor should be 1')
|
||||
|
||||
self.assertEqual(
|
||||
scale_factor,
|
||||
1,
|
||||
msg="Values are already in [0, 255], scale factor should be 1",
|
||||
)
|
||||
|
||||
def test_prepare_video(self):
|
||||
# At each timeframe, the sum over all other
|
||||
@ -183,7 +201,7 @@ class TestTensorBoardUtils(BaseTestCase):
|
||||
(16, 30, 3, 28, 28),
|
||||
(36, 30, 3, 28, 28),
|
||||
(19, 29, 3, 23, 19),
|
||||
(3, 3, 3, 3, 3)
|
||||
(3, 3, 3, 3, 3),
|
||||
]
|
||||
for s in shapes:
|
||||
V_input = np.random.random(s)
|
||||
@ -205,6 +223,7 @@ class TestTensorBoardUtils(BaseTestCase):
|
||||
y = np.reshape(V_after[f], newshape=(-1))
|
||||
np.testing.assert_array_almost_equal(np.sum(x), np.sum(y))
|
||||
|
||||
|
||||
freqs = [262, 294, 330, 349, 392, 440, 440, 440, 440, 440, 440]
|
||||
|
||||
true_positive_counts = [75, 64, 21, 5, 0]
|
||||
@ -214,6 +233,7 @@ false_negative_counts = [0, 11, 54, 70, 75]
|
||||
precision = [0.3333333, 0.3786982, 0.5384616, 1.0, 0.0]
|
||||
recall = [1.0, 0.8533334, 0.28, 0.0666667, 0.0]
|
||||
|
||||
|
||||
class TestTensorBoardWriter(BaseTestCase):
|
||||
def test_writer(self):
|
||||
with self.createSummaryWriter() as writer:
|
||||
@ -221,49 +241,66 @@ class TestTensorBoardWriter(BaseTestCase):
|
||||
|
||||
n_iter = 0
|
||||
writer.add_hparams(
|
||||
{'lr': 0.1, 'bsize': 1},
|
||||
{'hparam/accuracy': 10, 'hparam/loss': 10}
|
||||
{"lr": 0.1, "bsize": 1}, {"hparam/accuracy": 10, "hparam/loss": 10}
|
||||
)
|
||||
writer.add_scalar("data/scalar_systemtime", 0.1, n_iter)
|
||||
writer.add_scalar("data/scalar_customtime", 0.2, n_iter, walltime=n_iter)
|
||||
writer.add_scalar("data/new_style", 0.2, n_iter, new_style=True)
|
||||
writer.add_scalars(
|
||||
"data/scalar_group",
|
||||
{
|
||||
"xsinx": n_iter * np.sin(n_iter),
|
||||
"xcosx": n_iter * np.cos(n_iter),
|
||||
"arctanx": np.arctan(n_iter),
|
||||
},
|
||||
n_iter,
|
||||
)
|
||||
writer.add_scalar('data/scalar_systemtime', 0.1, n_iter)
|
||||
writer.add_scalar('data/scalar_customtime', 0.2, n_iter, walltime=n_iter)
|
||||
writer.add_scalar('data/new_style', 0.2, n_iter, new_style=True)
|
||||
writer.add_scalars('data/scalar_group', {
|
||||
"xsinx": n_iter * np.sin(n_iter),
|
||||
"xcosx": n_iter * np.cos(n_iter),
|
||||
"arctanx": np.arctan(n_iter)
|
||||
}, n_iter)
|
||||
x = np.zeros((32, 3, 64, 64)) # output from network
|
||||
writer.add_images('Image', x, n_iter) # Tensor
|
||||
writer.add_image_with_boxes('imagebox',
|
||||
np.zeros((3, 64, 64)),
|
||||
np.array([[10, 10, 40, 40], [40, 40, 60, 60]]),
|
||||
n_iter)
|
||||
writer.add_images("Image", x, n_iter) # Tensor
|
||||
writer.add_image_with_boxes(
|
||||
"imagebox",
|
||||
np.zeros((3, 64, 64)),
|
||||
np.array([[10, 10, 40, 40], [40, 40, 60, 60]]),
|
||||
n_iter,
|
||||
)
|
||||
x = np.zeros(sample_rate * 2)
|
||||
|
||||
writer.add_audio('myAudio', x, n_iter)
|
||||
writer.add_video('myVideo', np.random.rand(16, 48, 1, 28, 28).astype(np.float32), n_iter)
|
||||
writer.add_text('Text', 'text logged at step:' + str(n_iter), n_iter)
|
||||
writer.add_text('markdown Text', '''a|b\n-|-\nc|d''', n_iter)
|
||||
writer.add_histogram('hist', np.random.rand(100, 100), n_iter)
|
||||
writer.add_pr_curve('xoxo', np.random.randint(2, size=100), np.random.rand(
|
||||
100), n_iter) # needs tensorboard 0.4RC or later
|
||||
writer.add_pr_curve_raw('prcurve with raw data', true_positive_counts,
|
||||
false_positive_counts,
|
||||
true_negative_counts,
|
||||
false_negative_counts,
|
||||
precision,
|
||||
recall, n_iter)
|
||||
writer.add_audio("myAudio", x, n_iter)
|
||||
writer.add_video(
|
||||
"myVideo", np.random.rand(16, 48, 1, 28, 28).astype(np.float32), n_iter
|
||||
)
|
||||
writer.add_text("Text", "text logged at step:" + str(n_iter), n_iter)
|
||||
writer.add_text("markdown Text", """a|b\n-|-\nc|d""", n_iter)
|
||||
writer.add_histogram("hist", np.random.rand(100, 100), n_iter)
|
||||
writer.add_pr_curve(
|
||||
"xoxo", np.random.randint(2, size=100), np.random.rand(100), n_iter
|
||||
) # needs tensorboard 0.4RC or later
|
||||
writer.add_pr_curve_raw(
|
||||
"prcurve with raw data",
|
||||
true_positive_counts,
|
||||
false_positive_counts,
|
||||
true_negative_counts,
|
||||
false_negative_counts,
|
||||
precision,
|
||||
recall,
|
||||
n_iter,
|
||||
)
|
||||
|
||||
v = np.array([[[1, 1, 1], [-1, -1, 1], [1, -1, -1], [-1, 1, -1]]], dtype=float)
|
||||
c = np.array([[[255, 0, 0], [0, 255, 0], [0, 0, 255], [255, 0, 255]]], dtype=int)
|
||||
v = np.array(
|
||||
[[[1, 1, 1], [-1, -1, 1], [1, -1, -1], [-1, 1, -1]]], dtype=float
|
||||
)
|
||||
c = np.array(
|
||||
[[[255, 0, 0], [0, 255, 0], [0, 0, 255], [255, 0, 255]]], dtype=int
|
||||
)
|
||||
f = np.array([[[0, 2, 3], [0, 3, 1], [0, 1, 2], [1, 3, 2]]], dtype=int)
|
||||
writer.add_mesh('my_mesh', vertices=v, colors=c, faces=f)
|
||||
writer.add_mesh("my_mesh", vertices=v, colors=c, faces=f)
|
||||
|
||||
|
||||
class TestTensorBoardSummaryWriter(BaseTestCase):
|
||||
def test_summary_writer_ctx(self):
|
||||
# after using a SummaryWriter as a ctx it should be closed
|
||||
with self.createSummaryWriter() as writer:
|
||||
writer.add_scalar('test', 1)
|
||||
writer.add_scalar("test", 1)
|
||||
self.assertIs(writer.file_writer, None)
|
||||
|
||||
def test_summary_writer_close(self):
|
||||
@ -280,158 +317,208 @@ class TestTensorBoardSummaryWriter(BaseTestCase):
|
||||
|
||||
def test_pathlib(self):
|
||||
import pathlib
|
||||
p = pathlib.Path('./pathlibtest' + str(uuid.uuid4()))
|
||||
|
||||
p = pathlib.Path("./pathlibtest" + str(uuid.uuid4()))
|
||||
with SummaryWriter(p) as writer:
|
||||
writer.add_scalar('test', 1)
|
||||
writer.add_scalar("test", 1)
|
||||
import shutil
|
||||
|
||||
shutil.rmtree(str(p))
|
||||
|
||||
|
||||
class TestTensorBoardEmbedding(BaseTestCase):
|
||||
def test_embedding(self):
|
||||
w = self.createSummaryWriter()
|
||||
all_features = torch.tensor([[1., 2., 3.], [5., 4., 1.], [3., 7., 7.]])
|
||||
all_labels = torch.tensor([33., 44., 55.])
|
||||
all_features = torch.tensor([[1.0, 2.0, 3.0], [5.0, 4.0, 1.0], [3.0, 7.0, 7.0]])
|
||||
all_labels = torch.tensor([33.0, 44.0, 55.0])
|
||||
all_images = torch.zeros(3, 3, 5, 5)
|
||||
|
||||
w.add_embedding(all_features,
|
||||
metadata=all_labels,
|
||||
label_img=all_images,
|
||||
global_step=2)
|
||||
w.add_embedding(
|
||||
all_features, metadata=all_labels, label_img=all_images, global_step=2
|
||||
)
|
||||
|
||||
dataset_label = ['test'] * 2 + ['train'] * 2
|
||||
dataset_label = ["test"] * 2 + ["train"] * 2
|
||||
all_labels = list(zip(all_labels, dataset_label))
|
||||
w.add_embedding(all_features,
|
||||
metadata=all_labels,
|
||||
label_img=all_images,
|
||||
metadata_header=['digit', 'dataset'],
|
||||
global_step=2)
|
||||
w.add_embedding(
|
||||
all_features,
|
||||
metadata=all_labels,
|
||||
label_img=all_images,
|
||||
metadata_header=["digit", "dataset"],
|
||||
global_step=2,
|
||||
)
|
||||
# assert...
|
||||
|
||||
def test_embedding_64(self):
|
||||
w = self.createSummaryWriter()
|
||||
all_features = torch.tensor([[1., 2., 3.], [5., 4., 1.], [3., 7., 7.]])
|
||||
all_labels = torch.tensor([33., 44., 55.])
|
||||
all_features = torch.tensor([[1.0, 2.0, 3.0], [5.0, 4.0, 1.0], [3.0, 7.0, 7.0]])
|
||||
all_labels = torch.tensor([33.0, 44.0, 55.0])
|
||||
all_images = torch.zeros((3, 3, 5, 5), dtype=torch.float64)
|
||||
|
||||
w.add_embedding(all_features,
|
||||
metadata=all_labels,
|
||||
label_img=all_images,
|
||||
global_step=2)
|
||||
w.add_embedding(
|
||||
all_features, metadata=all_labels, label_img=all_images, global_step=2
|
||||
)
|
||||
|
||||
dataset_label = ['test'] * 2 + ['train'] * 2
|
||||
dataset_label = ["test"] * 2 + ["train"] * 2
|
||||
all_labels = list(zip(all_labels, dataset_label))
|
||||
w.add_embedding(all_features,
|
||||
metadata=all_labels,
|
||||
label_img=all_images,
|
||||
metadata_header=['digit', 'dataset'],
|
||||
global_step=2)
|
||||
w.add_embedding(
|
||||
all_features,
|
||||
metadata=all_labels,
|
||||
label_img=all_images,
|
||||
metadata_header=["digit", "dataset"],
|
||||
global_step=2,
|
||||
)
|
||||
|
||||
|
||||
class TestTensorBoardSummary(BaseTestCase):
|
||||
def test_uint8_image(self):
|
||||
'''
|
||||
"""
|
||||
Tests that uint8 image (pixel values in [0, 255]) is not changed
|
||||
'''
|
||||
"""
|
||||
test_image = np.random.randint(0, 256, size=(3, 32, 32), dtype=np.uint8)
|
||||
scale_factor = summary._calc_scale_factor(test_image)
|
||||
self.assertEqual(scale_factor, 1, msg='Values are already in [0, 255], scale factor should be 1')
|
||||
self.assertEqual(
|
||||
scale_factor,
|
||||
1,
|
||||
msg="Values are already in [0, 255], scale factor should be 1",
|
||||
)
|
||||
|
||||
def test_float32_image(self):
|
||||
'''
|
||||
"""
|
||||
Tests that float32 image (pixel values in [0, 1]) are scaled correctly
|
||||
to [0, 255]
|
||||
'''
|
||||
"""
|
||||
test_image = np.random.rand(3, 32, 32).astype(np.float32)
|
||||
scale_factor = summary._calc_scale_factor(test_image)
|
||||
self.assertEqual(scale_factor, 255, msg='Values are in [0, 1], scale factor should be 255')
|
||||
self.assertEqual(
|
||||
scale_factor, 255, msg="Values are in [0, 1], scale factor should be 255"
|
||||
)
|
||||
|
||||
def test_list_input(self):
|
||||
with self.assertRaises(Exception) as e_info:
|
||||
summary.histogram('dummy', [1, 3, 4, 5, 6], 'tensorflow')
|
||||
summary.histogram("dummy", [1, 3, 4, 5, 6], "tensorflow")
|
||||
|
||||
def test_empty_input(self):
|
||||
with self.assertRaises(Exception) as e_info:
|
||||
summary.histogram('dummy', np.ndarray(0), 'tensorflow')
|
||||
summary.histogram("dummy", np.ndarray(0), "tensorflow")
|
||||
|
||||
def test_image_with_boxes(self):
|
||||
self.assertTrue(compare_image_proto(summary.image_boxes('dummy',
|
||||
tensor_N(shape=(3, 32, 32)),
|
||||
np.array([[10, 10, 40, 40]])),
|
||||
self))
|
||||
self.assertTrue(
|
||||
compare_image_proto(
|
||||
summary.image_boxes(
|
||||
"dummy", tensor_N(shape=(3, 32, 32)), np.array([[10, 10, 40, 40]])
|
||||
),
|
||||
self,
|
||||
)
|
||||
)
|
||||
|
||||
def test_image_with_one_channel(self):
|
||||
self.assertTrue(compare_image_proto(
|
||||
summary.image('dummy',
|
||||
tensor_N(shape=(1, 8, 8)),
|
||||
dataformats='CHW'),
|
||||
self)) # noqa: E131
|
||||
self.assertTrue(
|
||||
compare_image_proto(
|
||||
summary.image("dummy", tensor_N(shape=(1, 8, 8)), dataformats="CHW"),
|
||||
self,
|
||||
)
|
||||
) # noqa: E131
|
||||
|
||||
def test_image_with_one_channel_batched(self):
|
||||
self.assertTrue(compare_image_proto(
|
||||
summary.image('dummy',
|
||||
tensor_N(shape=(2, 1, 8, 8)),
|
||||
dataformats='NCHW'),
|
||||
self)) # noqa: E131
|
||||
self.assertTrue(
|
||||
compare_image_proto(
|
||||
summary.image(
|
||||
"dummy", tensor_N(shape=(2, 1, 8, 8)), dataformats="NCHW"
|
||||
),
|
||||
self,
|
||||
)
|
||||
) # noqa: E131
|
||||
|
||||
def test_image_with_3_channel_batched(self):
|
||||
self.assertTrue(compare_image_proto(
|
||||
summary.image('dummy',
|
||||
tensor_N(shape=(2, 3, 8, 8)),
|
||||
dataformats='NCHW'),
|
||||
self)) # noqa: E131
|
||||
self.assertTrue(
|
||||
compare_image_proto(
|
||||
summary.image(
|
||||
"dummy", tensor_N(shape=(2, 3, 8, 8)), dataformats="NCHW"
|
||||
),
|
||||
self,
|
||||
)
|
||||
) # noqa: E131
|
||||
|
||||
def test_image_without_channel(self):
|
||||
self.assertTrue(compare_image_proto(
|
||||
summary.image('dummy',
|
||||
tensor_N(shape=(8, 8)),
|
||||
dataformats='HW'),
|
||||
self)) # noqa: E131
|
||||
self.assertTrue(
|
||||
compare_image_proto(
|
||||
summary.image("dummy", tensor_N(shape=(8, 8)), dataformats="HW"), self
|
||||
)
|
||||
) # noqa: E131
|
||||
|
||||
def test_video(self):
|
||||
try:
|
||||
import moviepy # noqa: F401
|
||||
except ImportError:
|
||||
return
|
||||
self.assertTrue(compare_proto(summary.video('dummy', tensor_N(shape=(4, 3, 1, 8, 8))), self))
|
||||
summary.video('dummy', np.random.rand(16, 48, 1, 28, 28))
|
||||
summary.video('dummy', np.random.rand(20, 7, 1, 8, 8))
|
||||
self.assertTrue(
|
||||
compare_proto(summary.video("dummy", tensor_N(shape=(4, 3, 1, 8, 8))), self)
|
||||
)
|
||||
summary.video("dummy", np.random.rand(16, 48, 1, 28, 28))
|
||||
summary.video("dummy", np.random.rand(20, 7, 1, 8, 8))
|
||||
|
||||
def test_audio(self):
|
||||
self.assertTrue(compare_proto(summary.audio('dummy', tensor_N(shape=(42,))), self))
|
||||
self.assertTrue(
|
||||
compare_proto(summary.audio("dummy", tensor_N(shape=(42,))), self)
|
||||
)
|
||||
|
||||
def test_text(self):
|
||||
self.assertTrue(compare_proto(summary.text('dummy', 'text 123'), self))
|
||||
self.assertTrue(compare_proto(summary.text("dummy", "text 123"), self))
|
||||
|
||||
def test_histogram_auto(self):
|
||||
self.assertTrue(compare_proto(summary.histogram('dummy', tensor_N(shape=(1024,)), bins='auto', max_bins=5), self))
|
||||
self.assertTrue(
|
||||
compare_proto(
|
||||
summary.histogram(
|
||||
"dummy", tensor_N(shape=(1024,)), bins="auto", max_bins=5
|
||||
),
|
||||
self,
|
||||
)
|
||||
)
|
||||
|
||||
def test_histogram_fd(self):
|
||||
self.assertTrue(compare_proto(summary.histogram('dummy', tensor_N(shape=(1024,)), bins='fd', max_bins=5), self))
|
||||
self.assertTrue(
|
||||
compare_proto(
|
||||
summary.histogram(
|
||||
"dummy", tensor_N(shape=(1024,)), bins="fd", max_bins=5
|
||||
),
|
||||
self,
|
||||
)
|
||||
)
|
||||
|
||||
def test_histogram_doane(self):
|
||||
self.assertTrue(compare_proto(summary.histogram('dummy', tensor_N(shape=(1024,)), bins='doane', max_bins=5), self))
|
||||
self.assertTrue(
|
||||
compare_proto(
|
||||
summary.histogram(
|
||||
"dummy", tensor_N(shape=(1024,)), bins="doane", max_bins=5
|
||||
),
|
||||
self,
|
||||
)
|
||||
)
|
||||
|
||||
def test_custom_scalars(self):
|
||||
layout = {
|
||||
'Taiwan': {
|
||||
'twse': ['Multiline', ['twse/0050', 'twse/2330']]
|
||||
"Taiwan": {"twse": ["Multiline", ["twse/0050", "twse/2330"]]},
|
||||
"USA": {
|
||||
"dow": ["Margin", ["dow/aaa", "dow/bbb", "dow/ccc"]],
|
||||
"nasdaq": ["Margin", ["nasdaq/aaa", "nasdaq/bbb", "nasdaq/ccc"]],
|
||||
},
|
||||
'USA': {
|
||||
'dow': ['Margin', ['dow/aaa', 'dow/bbb', 'dow/ccc']],
|
||||
'nasdaq': ['Margin', ['nasdaq/aaa', 'nasdaq/bbb', 'nasdaq/ccc']]
|
||||
}
|
||||
}
|
||||
summary.custom_scalars(layout) # only smoke test. Because protobuf in python2/3 serialize dictionary differently.
|
||||
summary.custom_scalars(
|
||||
layout
|
||||
) # only smoke test. Because protobuf in python2/3 serialize dictionary differently.
|
||||
|
||||
def test_hparams_smoke(self):
|
||||
hp = {'lr': 0.1, 'bsize': 4}
|
||||
mt = {'accuracy': 0.1, 'loss': 10}
|
||||
summary.hparams(hp, mt) # only smoke test. Because protobuf in python2/3 serialize dictionary differently.
|
||||
hp = {"lr": 0.1, "bsize": 4}
|
||||
mt = {"accuracy": 0.1, "loss": 10}
|
||||
summary.hparams(
|
||||
hp, mt
|
||||
) # only smoke test. Because protobuf in python2/3 serialize dictionary differently.
|
||||
|
||||
hp = {'use_magic': True, 'init_string': "42"}
|
||||
mt = {'accuracy': 0.1, 'loss': 10}
|
||||
hp = {"use_magic": True, "init_string": "42"}
|
||||
mt = {"accuracy": 0.1, "loss": 10}
|
||||
summary.hparams(hp, mt)
|
||||
|
||||
mt = {'accuracy': torch.zeros(1), 'loss': torch.zeros(1)}
|
||||
mt = {"accuracy": torch.zeros(1), "loss": torch.zeros(1)}
|
||||
summary.hparams(hp, mt)
|
||||
|
||||
def test_hparams_wrong_parameter(self):
|
||||
@ -440,25 +527,25 @@ class TestTensorBoardSummary(BaseTestCase):
|
||||
with self.assertRaises(TypeError):
|
||||
summary.hparams({}, [])
|
||||
with self.assertRaises(ValueError):
|
||||
res = summary.hparams({'pytorch': [1, 2]}, {'accuracy': 2.0})
|
||||
res = summary.hparams({"pytorch": [1, 2]}, {"accuracy": 2.0})
|
||||
# metric data is used in writer.py so the code path is different, which leads to different exception type.
|
||||
with self.assertRaises(NotImplementedError):
|
||||
with self.createSummaryWriter() as writer:
|
||||
writer.add_hparams({'pytorch': 1.0}, {'accuracy': [1, 2]})
|
||||
writer.add_hparams({"pytorch": 1.0}, {"accuracy": [1, 2]})
|
||||
|
||||
def test_hparams_number(self):
|
||||
hp = {'lr': 0.1}
|
||||
mt = {'accuracy': 0.1}
|
||||
hp = {"lr": 0.1}
|
||||
mt = {"accuracy": 0.1}
|
||||
self.assertTrue(compare_proto(summary.hparams(hp, mt), self))
|
||||
|
||||
def test_hparams_bool(self):
|
||||
hp = {'bool_var': True}
|
||||
mt = {'accuracy': 0.1}
|
||||
hp = {"bool_var": True}
|
||||
mt = {"accuracy": 0.1}
|
||||
self.assertTrue(compare_proto(summary.hparams(hp, mt), self))
|
||||
|
||||
def test_hparams_string(self):
|
||||
hp = {'string_var': "hi"}
|
||||
mt = {'accuracy': 0.1}
|
||||
hp = {"string_var": "hi"}
|
||||
mt = {"accuracy": 0.1}
|
||||
self.assertTrue(compare_proto(summary.hparams(hp, mt), self))
|
||||
|
||||
def test_hparams_domain_discrete(self):
|
||||
@ -479,30 +566,35 @@ class TestTensorBoardSummary(BaseTestCase):
|
||||
|
||||
def test_mesh(self):
|
||||
v = np.array([[[1, 1, 1], [-1, -1, 1], [1, -1, -1], [-1, 1, -1]]], dtype=float)
|
||||
c = np.array([[[255, 0, 0], [0, 255, 0], [0, 0, 255], [255, 0, 255]]], dtype=int)
|
||||
c = np.array(
|
||||
[[[255, 0, 0], [0, 255, 0], [0, 0, 255], [255, 0, 255]]], dtype=int
|
||||
)
|
||||
f = np.array([[[0, 2, 3], [0, 3, 1], [0, 1, 2], [1, 3, 2]]], dtype=int)
|
||||
mesh = summary.mesh('my_mesh', vertices=v, colors=c, faces=f, config_dict=None)
|
||||
mesh = summary.mesh("my_mesh", vertices=v, colors=c, faces=f, config_dict=None)
|
||||
self.assertTrue(compare_proto(mesh, self))
|
||||
|
||||
def test_scalar_new_style(self):
|
||||
scalar = summary.scalar('test_scalar', 1.0, new_style=True)
|
||||
scalar = summary.scalar("test_scalar", 1.0, new_style=True)
|
||||
self.assertTrue(compare_proto(scalar, self))
|
||||
|
||||
|
||||
def remove_whitespace(string):
|
||||
return string.replace(' ', '').replace('\t', '').replace('\n', '')
|
||||
return string.replace(" ", "").replace("\t", "").replace("\n", "")
|
||||
|
||||
|
||||
def get_expected_file(function_ptr):
|
||||
module_id = function_ptr.__class__.__module__
|
||||
test_file = sys.modules[module_id].__file__
|
||||
# Look for the .py file (since __file__ could be pyc).
|
||||
test_file = ".".join(test_file.split('.')[:-1]) + '.py'
|
||||
test_file = ".".join(test_file.split(".")[:-1]) + ".py"
|
||||
|
||||
# Use realpath to follow symlinks appropriately.
|
||||
test_dir = os.path.dirname(os.path.realpath(test_file))
|
||||
functionName = function_ptr.id().split('.')[-1]
|
||||
return os.path.join(test_dir,
|
||||
"expect",
|
||||
'TestTensorBoard.' + functionName + ".expect")
|
||||
functionName = function_ptr.id().split(".")[-1]
|
||||
return os.path.join(
|
||||
test_dir, "expect", "TestTensorBoard." + functionName + ".expect"
|
||||
)
|
||||
|
||||
|
||||
def read_expected_content(function_ptr):
|
||||
expected_file = get_expected_file(function_ptr)
|
||||
@ -510,6 +602,7 @@ def read_expected_content(function_ptr):
|
||||
with open(expected_file, "r") as f:
|
||||
return f.read()
|
||||
|
||||
|
||||
def compare_image_proto(actual_proto, function_ptr):
|
||||
expected_str = read_expected_content(function_ptr)
|
||||
expected_proto = Summary()
|
||||
@ -520,23 +613,26 @@ def compare_image_proto(actual_proto, function_ptr):
|
||||
expected_img = Image.open(io.BytesIO(expected.image.encoded_image_string))
|
||||
|
||||
return (
|
||||
actual.tag == expected.tag and
|
||||
actual.image.height == expected.image.height and
|
||||
actual.image.width == expected.image.width and
|
||||
actual.image.colorspace == expected.image.colorspace and
|
||||
actual_img == expected_img
|
||||
actual.tag == expected.tag
|
||||
and actual.image.height == expected.image.height
|
||||
and actual.image.width == expected.image.width
|
||||
and actual.image.colorspace == expected.image.colorspace
|
||||
and actual_img == expected_img
|
||||
)
|
||||
|
||||
|
||||
def compare_proto(str_to_compare, function_ptr):
|
||||
expected = read_expected_content(function_ptr)
|
||||
str_to_compare = str(str_to_compare)
|
||||
return remove_whitespace(str_to_compare) == remove_whitespace(expected)
|
||||
|
||||
|
||||
def write_proto(str_to_compare, function_ptr):
|
||||
expected_file = get_expected_file(function_ptr)
|
||||
with open(expected_file, 'w') as f:
|
||||
with open(expected_file, "w") as f:
|
||||
f.write(str(str_to_compare))
|
||||
|
||||
|
||||
class TestTensorBoardPytorchGraph(BaseTestCase):
|
||||
def test_pytorch_graph(self):
|
||||
dummy_input = (torch.zeros(1, 3),)
|
||||
@ -567,7 +663,8 @@ class TestTensorBoardPytorchGraph(BaseTestCase):
|
||||
self.assertEquals(expected_node.input, actual_node.input)
|
||||
self.assertEquals(expected_node.device, actual_node.device)
|
||||
self.assertEquals(
|
||||
sorted(expected_node.attr.keys()), sorted(actual_node.attr.keys()))
|
||||
sorted(expected_node.attr.keys()), sorted(actual_node.attr.keys())
|
||||
)
|
||||
|
||||
def test_mlp_graph(self):
|
||||
dummy_input = (torch.zeros(2, 1, 28, 28),)
|
||||
@ -587,8 +684,7 @@ class TestTensorBoardPytorchGraph(BaseTestCase):
|
||||
self.fc3 = torch.nn.Linear(1200, 10)
|
||||
|
||||
def forward(self, x, update_batch_stats=True):
|
||||
h = torch.nn.functional.relu(
|
||||
self.fc1(x.view(-1, self.input_len)))
|
||||
h = torch.nn.functional.relu(self.fc1(x.view(-1, self.input_len)))
|
||||
h = self.fc2(h)
|
||||
h = torch.nn.functional.relu(h)
|
||||
h = self.fc3(h)
|
||||
@ -607,42 +703,45 @@ class TestTensorBoardPytorchGraph(BaseTestCase):
|
||||
@skipIfNoTorchVision
|
||||
def test_torchvision_smoke(self):
|
||||
model_input_shapes = {
|
||||
'alexnet': (2, 3, 224, 224),
|
||||
'resnet34': (2, 3, 224, 224),
|
||||
'resnet152': (2, 3, 224, 224),
|
||||
'densenet121': (2, 3, 224, 224),
|
||||
'vgg16': (2, 3, 224, 224),
|
||||
'vgg19': (2, 3, 224, 224),
|
||||
'vgg16_bn': (2, 3, 224, 224),
|
||||
'vgg19_bn': (2, 3, 224, 224),
|
||||
'mobilenet_v2': (2, 3, 224, 224),
|
||||
"alexnet": (2, 3, 224, 224),
|
||||
"resnet34": (2, 3, 224, 224),
|
||||
"resnet152": (2, 3, 224, 224),
|
||||
"densenet121": (2, 3, 224, 224),
|
||||
"vgg16": (2, 3, 224, 224),
|
||||
"vgg19": (2, 3, 224, 224),
|
||||
"vgg16_bn": (2, 3, 224, 224),
|
||||
"vgg19_bn": (2, 3, 224, 224),
|
||||
"mobilenet_v2": (2, 3, 224, 224),
|
||||
}
|
||||
for model_name, input_shape in model_input_shapes.items():
|
||||
with self.createSummaryWriter() as w:
|
||||
model = getattr(torchvision.models, model_name)()
|
||||
w.add_graph(model, torch.zeros(input_shape))
|
||||
|
||||
|
||||
class TestTensorBoardFigure(BaseTestCase):
|
||||
@skipIfNoMatplotlib
|
||||
def test_figure(self):
|
||||
writer = self.createSummaryWriter()
|
||||
|
||||
figure, axes = plt.figure(), plt.gca()
|
||||
circle1 = plt.Circle((0.2, 0.5), 0.2, color='r')
|
||||
circle2 = plt.Circle((0.8, 0.5), 0.2, color='g')
|
||||
circle1 = plt.Circle((0.2, 0.5), 0.2, color="r")
|
||||
circle2 = plt.Circle((0.8, 0.5), 0.2, color="g")
|
||||
axes.add_patch(circle1)
|
||||
axes.add_patch(circle2)
|
||||
plt.axis('scaled')
|
||||
plt.axis("scaled")
|
||||
plt.tight_layout()
|
||||
|
||||
writer.add_figure("add_figure/figure", figure, 0, close=False)
|
||||
self.assertTrue(plt.fignum_exists(figure.number))
|
||||
|
||||
writer.add_figure("add_figure/figure", figure, 1)
|
||||
if matplotlib.__version__ != '3.3.0':
|
||||
if matplotlib.__version__ != "3.3.0":
|
||||
self.assertFalse(plt.fignum_exists(figure.number))
|
||||
else:
|
||||
print("Skipping fignum_exists, see https://github.com/matplotlib/matplotlib/issues/18163")
|
||||
print(
|
||||
"Skipping fignum_exists, see https://github.com/matplotlib/matplotlib/issues/18163"
|
||||
)
|
||||
|
||||
writer.close()
|
||||
|
||||
@ -661,16 +760,23 @@ class TestTensorBoardFigure(BaseTestCase):
|
||||
figures.append(figure)
|
||||
|
||||
writer.add_figure("add_figure/figure_list", figures, 0, close=False)
|
||||
self.assertTrue(all([plt.fignum_exists(figure.number) is True for figure in figures])) # noqa: F812
|
||||
self.assertTrue(
|
||||
all([plt.fignum_exists(figure.number) is True for figure in figures])
|
||||
) # noqa: F812
|
||||
|
||||
writer.add_figure("add_figure/figure_list", figures, 1)
|
||||
if matplotlib.__version__ != '3.3.0':
|
||||
self.assertTrue(all([plt.fignum_exists(figure.number) is False for figure in figures])) # noqa: F812
|
||||
if matplotlib.__version__ != "3.3.0":
|
||||
self.assertTrue(
|
||||
all([plt.fignum_exists(figure.number) is False for figure in figures])
|
||||
) # noqa: F812
|
||||
else:
|
||||
print("Skipping fignum_exists, see https://github.com/matplotlib/matplotlib/issues/18163")
|
||||
print(
|
||||
"Skipping fignum_exists, see https://github.com/matplotlib/matplotlib/issues/18163"
|
||||
)
|
||||
|
||||
writer.close()
|
||||
|
||||
|
||||
class TestTensorBoardNumpy(BaseTestCase):
|
||||
def test_scalar(self):
|
||||
res = make_np(1.1)
|
||||
@ -687,16 +793,16 @@ class TestTensorBoardNumpy(BaseTestCase):
|
||||
@skipIfNoCaffe2
|
||||
def test_caffe2_np(self):
|
||||
workspace.FeedBlob("testBlob", tensor_N(shape=(1, 3, 64, 64)))
|
||||
self.assertIsInstance(make_np('testBlob'), np.ndarray)
|
||||
self.assertIsInstance(make_np("testBlob"), np.ndarray)
|
||||
|
||||
@skipIfNoCaffe2
|
||||
def test_caffe2_np_expect_fail(self):
|
||||
with self.assertRaises(RuntimeError):
|
||||
res = make_np('This_blob_does_not_exist')
|
||||
res = make_np("This_blob_does_not_exist")
|
||||
|
||||
def test_pytorch_np_expect_fail(self):
|
||||
with self.assertRaises(NotImplementedError):
|
||||
res = make_np({'pytorch': 1.0})
|
||||
res = make_np({"pytorch": 1.0})
|
||||
|
||||
@skipIfNoCaffe2
|
||||
@unittest.skipIf(TEST_WITH_ASAN, "Caffe2 failure with ASAN")
|
||||
@ -707,20 +813,20 @@ class TestTensorBoardNumpy(BaseTestCase):
|
||||
workspace.FeedBlob("label", np.random.randn(1, 1000).astype(np.int))
|
||||
|
||||
with core.NameScope("conv1"):
|
||||
conv1 = brew.conv(model, "data", 'conv1', dim_in=1, dim_out=20, kernel=5)
|
||||
conv1 = brew.conv(model, "data", "conv1", dim_in=1, dim_out=20, kernel=5)
|
||||
# Image size: 24 x 24 -> 12 x 12
|
||||
pool1 = brew.max_pool(model, conv1, 'pool1', kernel=2, stride=2)
|
||||
pool1 = brew.max_pool(model, conv1, "pool1", kernel=2, stride=2)
|
||||
# Image size: 12 x 12 -> 8 x 8
|
||||
conv2 = brew.conv(model, pool1, 'conv2', dim_in=20, dim_out=100, kernel=5)
|
||||
conv2 = brew.conv(model, pool1, "conv2", dim_in=20, dim_out=100, kernel=5)
|
||||
# Image size: 8 x 8 -> 4 x 4
|
||||
pool2 = brew.max_pool(model, conv2, 'pool2', kernel=2, stride=2)
|
||||
pool2 = brew.max_pool(model, conv2, "pool2", kernel=2, stride=2)
|
||||
with core.NameScope("classifier"):
|
||||
# 50 * 4 * 4 stands for dim_out from previous layer multiplied by the image size
|
||||
fc3 = brew.fc(model, pool2, 'fc3', dim_in=100 * 4 * 4, dim_out=500)
|
||||
fc3 = brew.fc(model, pool2, "fc3", dim_in=100 * 4 * 4, dim_out=500)
|
||||
relu = brew.relu(model, fc3, fc3)
|
||||
pred = brew.fc(model, relu, 'pred', 500, 10)
|
||||
softmax = brew.softmax(model, pred, 'softmax')
|
||||
xent = model.LabelCrossEntropy([softmax, "label"], 'xent')
|
||||
pred = brew.fc(model, relu, "pred", 500, 10)
|
||||
softmax = brew.softmax(model, pred, "softmax")
|
||||
xent = model.LabelCrossEntropy([softmax, "label"], "xent")
|
||||
# compute the expected loss
|
||||
loss = model.AveragedLoss(xent, "loss")
|
||||
model.net.RunAllOnMKL()
|
||||
@ -759,5 +865,6 @@ class TestTensorBoardNumpy(BaseTestCase):
|
||||
)
|
||||
compare_proto(graph, self)
|
||||
|
||||
if __name__ == '__main__':
|
||||
|
||||
if __name__ == "__main__":
|
||||
run_tests()
|
||||
|
||||
Reference in New Issue
Block a user