def __init__(self, batch_size, num_threads, device_id, prefetch, seed):
super(CommonPipeline, self).__init__(batch_size,
num_threads,
device_id,
prefetch_queue_depth=prefetch)
self.decode_cpu = ops.HostDecoder(device="cpu", output_type=types.RGB)
self.decode_crop = ops.HostDecoderSlice(device="cpu",
output_type=types.RGB)
self.crop = ops.SSDRandomCrop(device="cpu", num_attempts=1, seed=seed)
self.crop2 = ops.RandomBBoxCrop(
device="cpu",
aspect_ratio=[0.5, 2.0],
thresholds=[0, 0.1, 0.3, 0.5, 0.7, 0.9],
scaling=[0.3, 1.0],
ltrb=True,
seed=seed)
self.slice_cpu = ops.Slice(device="cpu")
self.slice_gpu = ops.Slice(device="gpu")
self.flip_cpu = ops.Flip(device="cpu")
self.bb_flip_cpu = ops.BbFlip(device="cpu", ltrb=True)
self.flip_gpu = ops.Flip(device="gpu")
self.bb_flip_gpu = ops.BbFlip(device="gpu", ltrb=True)
def __init__(self, batch_size, num_threads, path, training, annotations, world, device_id, mean, std, resize, max_size, stride):
super().__init__(batch_size=batch_size, num_threads=num_threads, device_id = device_id, prefetch_queue_depth=num_threads, seed=42)
self.path = path
self.training = training
self.stride = stride
self.iter = 0
self.reader = ops.COCOReader(annotations_file=annotations, file_root=path, num_shards=world,shard_id=torch.cuda.current_device(),
ltrb=True, ratio=True, shuffle_after_epoch=True, save_img_ids=True)
self.decode_train = ops.ImageDecoderSlice(device="mixed", output_type=types.RGB)
self.decode_infer = ops.ImageDecoder(device="mixed", output_type=types.RGB)
self.bbox_crop = ops.RandomBBoxCrop(device='cpu', ltrb=True, scaling=[0.3, 1.0], thresholds=[0.1,0.3,0.5,0.7,0.9])
self.bbox_flip = ops.BbFlip(device='cpu', ltrb=True)
self.img_flip = ops.Flip(device='gpu')
self.coin_flip = ops.CoinFlip(probability=0.5)
if isinstance(resize, list): resize = max(resize)
self.rand_resize = ops.Uniform(range=[resize, float(max_size)])
self.resize_train = ops.Resize(device='gpu', interp_type=types.DALIInterpType.INTERP_CUBIC, save_attrs=True)
self.resize_infer = ops.Resize(device='gpu', interp_type=types.DALIInterpType.INTERP_CUBIC, resize_longer=max_size, save_attrs=True)
padded_size = max_size + ((self.stride - max_size % self.stride) % self.stride)
self.pad = ops.Paste(device='gpu', fill_value = 0, ratio=1.1, min_canvas_size=padded_size, paste_x=0, paste_y=0)
self.normalize = ops.CropMirrorNormalize(device='gpu', mean=mean, std=std, crop=padded_size, crop_pos_x=0, crop_pos_y=0)
def __init__(self, batch_size, num_threads, device_id):
super(TFRecordPipeline, self).__init__(batch_size,
num_threads,
device_id)
self.input = ops.TFRecordReader(path = tfrecord,
index_path = tfrecord_idx,
features = {"image/encoded" : tfrec.FixedLenFeature((), tfrec.string, ""),
'image/filename': tfrec.FixedLenFeature([ ], tfrec.string, ''),
'image/height': tfrec.FixedLenFeature([1], tfrec.int64, -1),
'image/width': tfrec.FixedLenFeature([1], tfrec.int64, -1),
'image/colorspace': tfrec.FixedLenFeature([ ], tfrec.string, ''),
'image/channels': tfrec.FixedLenFeature([1], tfrec.int64, -1),
'image/format': tfrec.FixedLenFeature([ ], tfrec.string, ''),
'image/class/label': tfrec.FixedLenFeature([1], tfrec.int64, -1),
'image/class/synset': tfrec.FixedLenFeature([ ], tfrec.string, ''),
'image/class/text': tfrec.FixedLenFeature([ ], tfrec.string, ''),
'image/object/bbox/xmin': tfrec.VarLenFeature(tfrec.float32, 0.0),
'image/object/bbox/ymin': tfrec.VarLenFeature(tfrec.float32, 0.0),
'image/object/bbox/xmax': tfrec.VarLenFeature(tfrec.float32, 0.0),
'image/object/bbox/ymax': tfrec.VarLenFeature(tfrec.float32, 0.0),
'image/object/bbox/label': tfrec.FixedLenFeature([1], tfrec.int64, -1)})
self.decode = ops.ImageDecoder(device = "mixed", output_type = types.RGB)
self.resize = ops.Resize(device = "gpu", resize_x = 1920.,resize_y=1920.)
self.vert_flip = ops.Flip(device = "gpu", horizontal=0)
self.vert_coin = ops.CoinFlip(probability=0.5)
self.cmnp = ops.CropMirrorNormalize(device = "gpu",
output_dtype = types.FLOAT,
crop = (1920, 1920),
image_type = types.RGB,
mean = [0., 0., 0.],
std = [1., 1., 1.])
self.uniform = ops.Uniform(range = (0.0, 1.0))
self.iter = 0
def __init__(self,
batch_size,
num_threads,
device_id,
data_dir,
crop,
size,
local_rank=0,
world_size=1):
super(HybridValPipe_CIFAR, self).__init__(batch_size,
num_threads,
device_id,
seed=12 + device_id)
self.iterator = iter(CIFAR_INPUT_ITER(batch_size, 'val',
root=data_dir))
self.input = ops.ExternalSource()
self.input_label = ops.ExternalSource()
self.pad = ops.Paste(device="gpu", ratio=1., fill_value=0)
self.uniform = ops.Uniform(range=(0., 1.))
self.crop = ops.Crop(device="gpu", crop_h=32, crop_w=32)
self.coin = ops.CoinFlip(probability=0.5)
self.flip = ops.Flip(device="gpu")
self.cmnp = ops.CropMirrorNormalize(device="gpu",
output_layout=types.NCHW,
mean=[125.31, 122.95, 113.87],
std=[63.0, 62.09, 66.70])
def __init__(self, params, device_id, files, labels):
super().__init__(params.batch_size,
params.num_gpus * 8,
device_id,
seed=params.seed)
# file_root有坑，并不是文件夹名字就是label，按照文件夹顺序(1, 10, 11, 2, 20, 21, ...)分别给与0，1，2，3，4...标签
self.input = ops.FileReader(files=files,
labels=labels,
random_shuffle=True)
self.decocer = ops.ImageDecoder(device='mixed', output_type=types.RGB)
self.resize = ops.Resize(device='gpu', resize_shorter=224)
self.pos_rng_x = ops.random.Uniform(range=(0.0, 1.0))
self.pos_rng_y = ops.random.Uniform(range=(0.0, 1.0))
self.crop = ops.Crop(device='gpu', crop_h=224, crop_w=224)
self.flip = ops.Flip(device='gpu')
self.coinflip = ops.random.CoinFlip(probability=0.5)
self.hsv = ops.Hsv(device='gpu')
self.saturation = ops.random.Uniform(range=(0.8, 1.0))
self.value = ops.random.Uniform(range=(0.8, 1.0))
mean = torch.Tensor(params.mean).unsqueeze(0).unsqueeze(0) * 255
std = torch.Tensor(params.std).unsqueeze(0).unsqueeze(0) * 255
self.normalize = ops.Normalize(axes=[0, 1],
mean=mean,
stddev=std,
device='gpu',
batch=False)
self.transpose = ops.Transpose(device='gpu', perm=[2, 0, 1])
def __new__(cls, horizontal=None, vertical=None, depthwise=None, **kwargs):
"""Create a ``Flip`` operator.
Parameters
----------
horizontal : int, optional
Whether to apply the horizontal flip.
vertical : int, optional
Whether to apply the vertical flip.
depthwise : bool, optional, default=True
Whether to apply the depthwise flip.
Returns
-------
nvidia.dali.ops.Flip
The operator.
"""
return ops.Flip(
horizontal=horizontal,
vertical=vertical,
depthwise=depthwise,
device=context.get_device_type(),
**kwargs
)
def __init__(self,
root,
split,
batch_size,
device_id,
dali_cpu=False,
local_rank=0,
world_size=1,
num_workers=2,
augment=False,
resize=False,
resize_size=314,
crop=True,
crop_size=314,
shuffle=True,
flip=True,
rotate=False,
rotate_angle=10.0):
super(VOCDali, self).__init__(batch_size,
num_threads=num_workers,
device_id=device_id,
seed=12 + device_id)
self.iterator = iter(VOCIter(batch_size, split, root, shuffle))
self.split = split
assert self.split in ['train', 'val']
dali_device = "gpu" if not dali_cpu else "cpu"
self.input = ops.ExternalSource()
self.input_label = ops.ExternalSource()
self.is_flip = flip
self.is_rotate = rotate
self.is_augment = augment
self.is_crop = crop
self.is_resize = resize
self.decode = ops.ImageDecoder(device='mixed', output_type=types.RGB)
# self.cast = ops.Cast(device='gpu', dtype=types.INT32)
self.rng = ops.Uniform(range=(0., 1.))
self.coin = ops.CoinFlip(probability=0.5)
# 定义大小
self.resize = ops.Resize(device="gpu",
resize_x=resize_size,
resize_y=resize_size,
interp_type=types.INTERP_TRIANGULAR)
# 定义旋转
self.rotate = ops.Rotate(device="gpu", angle=rotate_angle)
# 定义翻转
self.flip = ops.Flip(device="gpu", vertical=1, horizontal=0)
# 定义剪裁
self.crop = ops.Crop(device=dali_device,
crop_h=crop_size,
crop_w=crop_size)
# 定义正则化
self.cmnp = ops.CropMirrorNormalize(
device=dali_device,
output_dtype=types.FLOAT,
output_layout=types.NCHW,
image_type=types.RGB,
mean=[0.45734706 * 255, 0.43338275 * 255, 0.40058118 * 255],
std=[0.23965294 * 255, 0.23532275 * 255, 0.2398498 * 255])
def __init__(self,
DATA_PATH,
input_height,
batch_size,
copies,
stage,
num_threads,
device_id,
seed=1729):
super(SimCLRTransform, self).__init__(batch_size,
num_threads,
device_id,
seed=seed)
#this lets our pytorch compat function find the length of our dataset
self.num_samples = len(ImageFolder(DATA_PATH))
self.copies = copies
self.input_height = input_height
self.stage = stage
self.input = ops.FileReader(file_root=DATA_PATH,
random_shuffle=True,
seed=seed)
self.to_int64 = ops.Cast(dtype=types.INT64, device="gpu")
self.to_int32_cpu = ops.Cast(dtype=types.INT32, device="cpu")
self.coin = ops.random.CoinFlip(probability=0.5)
self.uniform = ops.random.Uniform(range=[0.6, 0.9])
self.blur_amt = ops.random.Uniform(values=[
float(i) for i in range(1, int(0.1 * self.input_height), 2)
])
self.angles = ops.random.Uniform(range=[0, 360])
self.cast = ops.Cast(dtype=types.FLOAT, device='gpu')
self.decode = ops.ImageDecoder(device='mixed', output_type=types.RGB)
self.crop = ops.RandomResizedCrop(size=self.input_height,
minibatch_size=batch_size,
random_area=[0.75, 1.0],
device="gpu")
self.resize = ops.Resize(resize_x=self.input_height,
resize_y=self.input_height,
device="gpu")
self.flip = ops.Flip(vertical=self.coin(),
horizontal=self.coin(),
device="gpu")
self.colorjit_gray = ops.ColorTwist(brightness=self.uniform(),
contrast=self.uniform(),
hue=self.uniform(),
saturation=self.uniform(),
device="gpu")
self.blur = ops.GaussianBlur(window_size=self.to_int32_cpu(
self.blur_amt()),
device="gpu")
self.rotate = ops.Rotate(
angle=self.angles(),
keep_size=True,
interp_type=types.DALIInterpType.INTERP_LINEAR,
device="gpu")
self.swapaxes = ops.Transpose(perm=[2, 0, 1], device="gpu")
def __init__(self, root_dir, batch_size, num_threads, device_id,
use_shift_scale=False,
num_shards=None, shard_id=None):
super().__init__(batch_size, num_threads, device_id, seed=12)
self.random_angle = ops.Uniform(range=(0, 360.0))
self.random = ops.Uniform(range=(0.5, 1.5))
self.random_coin = ops.CoinFlip()
self.input = ops.FileReader(
file_root=root_dir, random_shuffle=True,
num_shards=num_shards, shard_id=shard_id,
)
self.decode = ops.ImageDecoder(device='mixed')
self.rotate = ops.Rotate(device='gpu', interp_type=types.INTERP_LINEAR)
self.crop = ops.Crop(device='gpu', crop=(224, 224))
self.use_shift_scale = use_shift_scale
if self.use_shift_scale:
self.shift_scale = ops.RandomResizedCrop(
device='gpu',
size=(224, 224),
interp_type=types.INTERP_LINEAR,
random_area=(0.3, 1.0),
)
self.flip = ops.Flip(device='gpu')
self.color_twist = ops.ColorTwist(device='gpu')
def __init__(self, DATA_PATH, input_height, batch_size, num_threads,
device_id):
super(SimCLRTrainDataTransform, self).__init__(batch_size,
num_threads,
device_id,
seed=12)
self.COPIES = 3
self.input_height = input_height
self.input = ops.FileReader(file_root=DATA_PATH,
random_shuffle=True,
seed=12)
self.coin = ops.CoinFlip(probability=0.5)
self.uniform = ops.Uniform(range=[0.7, 1.3]) #-1 to 1
#read image (I think that has to be cpu, do a mixed operation to decode into gpu)
self.decode = ops.ImageDecoder(device='mixed', output_type=types.RGB)
self.crop = ops.RandomResizedCrop(size=self.input_height, device="gpu")
self.flip = ops.Flip(vertical=self.coin(),
horizontal=self.coin(),
device="gpu")
self.colorjit_gray = ops.ColorTwist(brightness=self.uniform(),
contrast=self.uniform(),
hue=self.uniform(),
saturation=self.uniform(),
device="gpu")
self.blur = ops.GaussianBlur(window_size=int(0.1 * self.input_height),
device="gpu")
self.swapaxes = ops.Transpose(perm=[2, 0, 1], device="gpu")
self.to_int64 = ops.Cast(dtype=types.INT64, device="gpu")
def __init__(self, batch_size, sequence_length, num_threads, device_id, file_root, crop_size, transforms=None):
super(VideoReaderPipeline, self).__init__(batch_size, num_threads, device_id, seed=12)
self.reader = ops.VideoReader(
device='gpu',
file_root=file_root,
sequence_length=sequence_length,
normalized=False,
random_shuffle=True,
image_type=types.RGB,
dtype=types.UINT8,
initial_fill=16
)
self.crop = ops.Crop(device="gpu", crop=crop_size, output_dtype=types.FLOAT)
self.transpose = ops.Transpose(device="gpu", perm=[3, 0, 1, 2])
self.uniform = ops.Uniform(range=(0.0, 1.0))
self.flip = ops.Flip(device="gpu", horizontal=1, vertical=0)
# self.normalize = ops.NormalizePermute(
# device="gpu",
# mean=[0.485, 0.456, 0.406],
# std=[0.229, 0.224, 0.225],
# width=224,
# height=224
# )
self.cmn = ops.CropMirrorNormalize(
device="gpu",
output_dtype=types.FLOAT,
# # output_layout=types.NCHW,
crop=(224, 224),
image_type=types.RGB,
mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225]
)
def __init__(self,
batch_size,
num_threads,
device_id,
data_dir,
crop,
size=256,
dali_cpu=False,
local_rank=0,
world_size=1):
super(HybridTrainPipe, self).__init__(batch_size,
num_threads,
device_id,
seed=12 + device_id)
dali_device = "gpu"
self.input = ops.FileReader(file_root=data_dir,
shard_id=local_rank,
num_shards=world_size,
random_shuffle=True)
self.decode = ops.ImageDecoder(device="mixed", output_type=types.RGB)
# self.res = ops.Resize(device="gpu", resize_x=size, resize_y=size, interp_type=types.INTERP_LINEAR)
self.res = ops.Resize(device="gpu",
resize_shorter=size,
interp_type=types.INTERP_LINEAR)
self.rescrop = ops.RandomResizedCrop(device="gpu",
size=crop,
random_area=[0.08, 1.25])
self.bc = ops.BrightnessContrast(device="gpu",
brightness=0.5,
contrast=0.6)
# Will flip vertically with prob of 0.1
self.vert_flip = ops.Flip(device='gpu', horizontal=0)
self.vert_coin = ops.CoinFlip(probability=0.4)
self.transform_source = ops.ExternalSource()
self.warp_keep_size = ops.WarpAffine(
device="gpu",
# size # keep original canvas size
interp_type=types.INTERP_LINEAR # use linear interpolation
)
# My workaround for Dali not supporting random affine transforms:
# a "synthetic random" warp affine transform.
# Rotate within a narrow range with probability of 0.075
self.rotate = ops.Rotate(device='gpu')
self.rotate_range = ops.Uniform(range=(-20.0, 20.0))
self.rotate_coin = ops.CoinFlip(probability=0.075)
self.cmnp = ops.CropMirrorNormalize(
device="gpu",
output_dtype=types.FLOAT,
output_layout=types.NCHW,
image_type=types.RGB,
mean=[0.485 * 255, 0.456 * 255, 0.406 * 255],
std=[0.229 * 255, 0.224 * 255, 0.225 * 255])
self.coin = ops.CoinFlip(probability=0.5)
print('DALI "{0}" variant'.format(dali_device))
def __init__(self, batch_size, layout, data_iterator, device):
super(SynthFlipPipeline, self).__init__(batch_size, seed=1234, num_threads=4, device_id=0)
self.device = device
self.iterator = data_iterator
self.layout = layout
self.input = ops.ExternalSource()
self.coin = ops.CoinFlip(seed=1234)
self.flip = ops.Flip(device=device)
def __init__(self, device):
super(CommonPipeline, self).__init__(BATCH_SIZE, NUM_WORKERS, DEVICE_ID, seed=SEED,
exec_async=False, exec_pipelined=False)
self.input = ops.FileReader(file_root=images_dir)
self.decode = ops.ImageDecoder(device='mixed' if device == 'gpu' else 'cpu',
output_type=types.RGB)
self.resize = ops.Resize(resize_x=400, resize_y=400, device=device)
self.flip = ops.Flip(device=device)
def __init__(self, device, batch_size, num_threads=1, device_id=0, num_gpus=1, is_vertical=0, is_horizontal=1 ):
super(FlipPipeline, self).__init__(batch_size,
num_threads,
device_id)
self.device = device
self.input = ops.CaffeReader(path = caffe_db_folder, shard_id = device_id, num_shards = num_gpus)
self.decode = ops.ImageDecoder(device = "cpu", output_type = types.RGB)
self.flip = ops.Flip(device = self.device, vertical=is_vertical, horizontal=is_horizontal)
def __init__(self,
batch_size,
device,
data_dir,
mean,
std,
device_id=0,
shard_id=0,
num_shards=1,
num_threads=4,
seed=0):
super(DaliTransformsTrainPipeline,
self).__init__(batch_size, num_threads, device_id, seed)
# should we make this drive the device flags?
self.reader = ops.FileReader(file_root=data_dir,
shard_id=shard_id,
num_shards=num_shards,
random_shuffle=True)
self.decode = ops.ImageDecoder(device='mixed',
output_type=types.RGB,
memory_stats=True)
self.resize = ops.Resize(device=device,
size=[200, 300],
interp_type=types.INTERP_TRIANGULAR)
self.centrecrop = ops.Crop(device=device, crop=[100, 100])
self.randomcrop = ops.RandomResizedCrop(device=device, size=[80, 80])
self.hz_coin = ops.CoinFlip(probability=0.5)
self.horizontalflip = ops.Flip(device=device)
self.rotate_angle = ops.Uniform(range=[-90, 90])
self.rotate_coin = ops.CoinFlip(probability=0.5)
self.rotate = ops.Rotate(device=device, keep_size=True)
self.vt_coin = ops.CoinFlip(probability=0.5)
self.verticalflip = ops.Flip(device=device, horizontal=0)
self.normalize = ops.CropMirrorNormalize(device=device,
dtype=types.FLOAT,
output_layout=types.NCHW) #,
#mean=mean*255, std=std*255)
#self.normalize = ops.Normalize(device=device, dtype=types.FLOAT)#, mean=mean, stddev=std)
self.to_int64 = ops.Cast(dtype=types.INT64, device=device)
def __init__(self, target_size, preproc_param, training=False):
self.training = training
mean = preproc_param.MEAN
std = preproc_param.STD
bri_delta = preproc_param.BRI_DELTA
hue_delta = preproc_param.HUE_DELTA
max_expand_ratio = preproc_param.MAX_EXPAND_RATIO
contrast_range = preproc_param.CONTRAST_RANGE
saturation_range = preproc_param.SATURATION_RANGE
crop_aspect_ratio = preproc_param.CROP_ASPECT_RATIO
crop_scale = preproc_param.CROP_SCALE
crop_attempts = preproc_param.CROP_ATTEMPTS
# decoder
self.decode_train = ops.ImageDecoderSlice(device="mixed", output_type=types.RGB)
self.decode_infer = ops.ImageDecoder(device="mixed", output_type=types.RGB)
# ssd crop
self.bbox_crop = ops.RandomBBoxCrop(
device="cpu",
bbox_layout="xyXY",
scaling=crop_scale,
aspect_ratio=crop_aspect_ratio,
allow_no_crop=True,
thresholds=[0, 0.1, 0.3, 0.5, 0.7, 0.9],
num_attempts=crop_attempts,
)
# color twist
self.uniform_con = ops.Uniform(range=contrast_range)
self.uniform_bri = ops.Uniform(
range=(1.0 - bri_delta / 256.0, 1.0 + bri_delta / 256.0)
)
self.uniform_sat = ops.Uniform(range=saturation_range)
self.uniform_hue = ops.Uniform(range=(-hue_delta, hue_delta))
self.hsv = ops.Hsv(device="gpu")
self.contrast = ops.BrightnessContrast(device="gpu")
# hflip
self.bbox_flip = ops.BbFlip(device="cpu", ltrb=True)
self.img_flip = ops.Flip(device="gpu")
self.coin_flip = ops.CoinFlip(probability=0.5)
# past
self.paste_pos = ops.Uniform(range=(0, 1))
self.paste_ratio = ops.Uniform(range=(1, max_expand_ratio))
self.paste = ops.Paste(device="gpu", fill_value=mean)
self.bbox_paste = ops.BBoxPaste(device="cpu", ltrb=True)
# resize and normalize
self.resize = ops.Resize(
device="gpu",
interp_type=types.DALIInterpType.INTERP_CUBIC,
resize_x=target_size[0],
resize_y=target_size[1],
save_attrs=True,
)
self.normalize = ops.CropMirrorNormalize(device="gpu", mean=mean, std=std)
def __init__(self,
imageset_dir,
image_size,
random_shuffle,
batch_size=4,
num_threads=2,
device_id=0):
super(ImagePipeline, self).__init__(batch_size,
num_threads,
device_id,
seed=12)
self.imageset_dir = imageset_dir
self.random_shuffle = random_shuffle
eii = ExternalInputIterator(root=self.imageset_dir,
batch_size=self.batch_size,
random_shuffle=self.random_shuffle)
self.iterator = iter(eii)
self.num_inputs = len(self.iterator.files)
self.input_image = ops.ExternalSource()
self.input_mask = ops.ExternalSource()
self.decode_image = ops.ImageDecoder(device="mixed",
output_type=types.RGB)
self.decode_mask = ops.ImageDecoder(device="mixed",
output_type=types.GRAY)
# The rest of pre-processing is done on the GPU
self.res = ops.Resize(device="gpu",
resize_x=image_size,
resize_y=image_size)
self.flip = ops.Flip(device="gpu", horizontal=1, vertical=0)
rotate_degree = random.random() * 2 * 10 - 10
self.rotate_image = ops.Rotate(
device="gpu",
angle=rotate_degree,
interp_type=types.DALIInterpType.INTERP_LINEAR)
self.rotate_mask = ops.Rotate(
device="gpu",
angle=rotate_degree,
interp_type=types.DALIInterpType.INTERP_NN)
self.cmnp_image = ops.CropMirrorNormalize(
device="gpu",
output_dtype=types.FLOAT,
output_layout=types.NCHW,
image_type=types.RGB,
mean=[0.485 * 255, 0.456 * 255, 0.406 * 255],
std=[0.229 * 255, 0.224 * 255, 0.225 * 255])
self.cmnp_mask = ops.CropMirrorNormalize(device="gpu",
output_dtype=types.FLOAT,
output_layout=types.NCHW,
image_type=types.GRAY,
mean=[0],
std=[255])
def __init__(self, batch_size, num_threads, device_id, data_dir, crop, dali_cpu=False):
super(HybridTrainPipe, self).__init__(batch_size, num_threads, device_id, seed=12 + device_id)
self.input = ops.FileReader(file_root=data_dir, shard_id=args.local_rank, num_shards=args.world_size, random_shuffle=True)
#let user decide which pipeline works him bets for RN version he runs
dali_device = 'cpu' if dali_cpu else 'gpu'
decoder_device = 'cpu' if dali_cpu else 'mixed'
# This padding sets the size of the internal nvJPEG buffers to be able to handle all images from full-sized ImageNet
# without additional reallocations
device_memory_padding = 211025920 if decoder_device == 'mixed' else 0
host_memory_padding = 140544512 if decoder_device == 'mixed' else 0
## randomly crop and resize, crop sampling,
self.decode = ops.ImageDecoderRandomCrop(device=decoder_device, output_type=types.RGB,
device_memory_padding=device_memory_padding,
host_memory_padding=host_memory_padding,
random_aspect_ratio=[0.75, 4/3.0],
random_area=[0.08, 1.0],
num_attempts=10)
self.res = ops.Resize(device=dali_device, resize_x=crop, resize_y=crop, interp_type=types.INTERP_TRIANGULAR)
self.vert_flip=ops.Flip(device='gpu', horizontal=0)
self.vert_coin = ops.CoinFlip(probability=0.075)
##color jitter https://www.gitmemory.com/ruiyuanlu, https://github.com/NVIDIA/DALI/issues/336
self.twist = ops.ColorTwist(device="gpu")
self.rng1 = ops.Uniform(range=[0.6, 1.4])
self.rng2 = ops.Uniform(range=[-102, 102]) ## factor=0.4, 0.4*255, -0.4*255
#self.rng2 = ops.Uniform(range=[-51, 51]) ## factor=0.2, 0.2*255, -0.2*255
self.flip = ops.Flip(device = "gpu", vertical = 1, horizontal = 0)
self.color_jitter = ops.ColorTwist(device="gpu", hue=0.2, brightness=0.4,
contrast=0.4, saturation=0.4)
self.cmnp = ops.CropMirrorNormalize(device="gpu",
output_dtype=types.FLOAT,
output_layout=types.NCHW,
crop=(crop, crop),
image_type=types.RGB,
mean=[0.485 * 255,0.456 * 255,0.406 * 255],
std=[0.229 * 255,0.224 * 255,0.225 * 255])
## this is torch.transform.RandomHorizontalFlip
self.mirrorcoin = ops.CoinFlip(probability=0.5)
self.uniform = ops.Uniform(range = (0.0, 1.0))
print('DALI "{0}" variant'.format(dali_device))
def __init__(self, p: float = 0.5):
"""Initialization
Args:
p (float, optional): Probability to apply this transformation. Defaults to 0.5.
"""
self.flip_coin_vflip = ops.CoinFlip(probability=p)
self.image_vflip = ops.Flip(device='gpu', horizontal=0)
self.bbox_vflip = ops.BbFlip(device='cpu', horizontal=0)
self.ldmrks_vflip = ops.CoordFlip(layout='xy', device='cpu', flip_x=0)
def __init__(self, batch_size, num_threads, path, training, annotations, world, device_id, mean, std, resize,
max_size, stride, rotate_augment=False,
augment_brightness=0.0,
augment_contrast=0.0, augment_hue=0.0,
augment_saturation=0.0):
super().__init__(batch_size=batch_size, num_threads=num_threads, device_id=device_id,
prefetch_queue_depth=num_threads, seed=42)
self.path = path
self.training = training
self.stride = stride
self.iter = 0
self.rotate_augment = rotate_augment
self.augment_brightness = augment_brightness
self.augment_contrast = augment_contrast
self.augment_hue = augment_hue
self.augment_saturation = augment_saturation
self.reader = ops.COCOReader(annotations_file=annotations, file_root=path, num_shards=world,
shard_id=torch.cuda.current_device(),
ltrb=True, ratio=True, shuffle_after_epoch=True, save_img_ids=True)
self.decode_train = ops.ImageDecoderSlice(device="mixed", output_type=types.RGB)
self.decode_infer = ops.ImageDecoder(device="mixed", output_type=types.RGB)
self.bbox_crop = ops.RandomBBoxCrop(device='cpu', ltrb=True, scaling=[0.3, 1.0],
thresholds=[0.1, 0.3, 0.5, 0.7, 0.9])
self.bbox_flip = ops.BbFlip(device='cpu', ltrb=True)
self.img_flip = ops.Flip(device='gpu')
self.coin_flip = ops.CoinFlip(probability=0.5)
self.bc = ops.BrightnessContrast(device='gpu')
self.hsv = ops.Hsv(device='gpu')
# Random number generation for augmentation
self.brightness_dist = ops.NormalDistribution(mean=1.0, stddev=augment_brightness)
self.contrast_dist = ops.NormalDistribution(mean=1.0, stddev=augment_contrast)
self.hue_dist = ops.NormalDistribution(mean=0.0, stddev=augment_hue)
self.saturation_dist = ops.NormalDistribution(mean=1.0, stddev=augment_saturation)
if rotate_augment:
raise RuntimeWarning("--augment-rotate current has no effect when using the DALI data loader.")
if isinstance(resize, list): resize = max(resize)
self.rand_resize = ops.Uniform(range=[resize, float(max_size)])
self.resize_train = ops.Resize(device='gpu', interp_type=types.DALIInterpType.INTERP_CUBIC, save_attrs=True)
self.resize_infer = ops.Resize(device='gpu', interp_type=types.DALIInterpType.INTERP_CUBIC,
resize_longer=max_size, save_attrs=True)
padded_size = max_size + ((self.stride - max_size % self.stride) % self.stride)
self.pad = ops.Paste(device='gpu', fill_value=0, ratio=1.1, min_canvas_size=padded_size, paste_x=0, paste_y=0)
self.normalize = ops.CropMirrorNormalize(device='gpu', mean=mean, std=std, crop=(padded_size, padded_size),
crop_pos_x=0, crop_pos_y=0)
def __init__(self, batch_size, num_threads, device_id, data_dir, crop,
mean, std, local_rank=0, world_size=1, dali_cpu=False, shuffle=True, fp16=False,
min_crop_size=0.08, color_jitter=False):
# As we're recreating the Pipeline at every epoch, the seed must be -1 (random seed)
super(HybridTrainPipe, self).__init__(
batch_size, num_threads, device_id, seed=-1)
# Enabling read_ahead slowed down processing ~40%
self.input = ops.FileReader(file_root=data_dir, shard_id=local_rank, num_shards=world_size,
random_shuffle=shuffle)
# Let user decide which pipeline works best with the chosen model
if dali_cpu:
decode_device = "cpu"
self.dali_device = "cpu"
self.flip = ops.Flip(device=self.dali_device)
else:
decode_device = "mixed"
self.dali_device = "gpu"
output_dtype = types.FLOAT
if fp16:
output_dtype = types.FLOAT16
self.cmn = ops.CropMirrorNormalize(device=self.dali_device,
output_dtype=output_dtype,
output_layout=types.NCHW,
crop=(crop, crop),
image_type=types.RGB,
mean=mean,
std=std,)
# To be able to handle all images from full-sized ImageNet, this padding sets the size of the internal
# nvJPEG buffers without additional reallocations
device_memory_padding = 211025920 if decode_device == 'mixed' else 0
host_memory_padding = 140544512 if decode_device == 'mixed' else 0
self.decode = ops.ImageDecoderRandomCrop(device=decode_device, output_type=types.RGB,
device_memory_padding=device_memory_padding,
host_memory_padding=host_memory_padding,
random_aspect_ratio=[
0.8, 1.25],
random_area=[
min_crop_size, 1.0],
num_attempts=100)
# Resize as desired. To match torchvision data loader, use triangular interpolation.
self.res = ops.Resize(device=self.dali_device, resize_x=crop, resize_y=crop,
interp_type=types.INTERP_TRIANGULAR)
self.coin = ops.CoinFlip(probability=0.5)
print('DALI "{0}" variant'.format(self.dali_device))
def __init__(self, data_iter, batch_size, x_dim, num_threads, device_id):
super(MiniPipeline, self).__init__(batch_size, num_threads, device_id, seed=12)
data_iter = iter(data_iter)
self.source = ops.ExternalSource(source=data_iter, num_outputs=2)
self.decode = ops.ImageDecoder(device='mixed', output_type=types.RGB)
self.rrc = ops.RandomResizedCrop(device='gpu', size=x_dim[:2], random_area=[0.95, 1.0])
self.flip = ops.Flip(device='gpu')
# self.colortwist = ops.ColorTwist(device='gpu', brightness=.1, contrast=.1, saturation=.1, hue=.1)
self.norm = ops.CropMirrorNormalize(device='gpu',
output_layout=types.NCHW,
# mean=[0.485 * 255, 0.456 * 255, 0.406 * 255],
# std=[0.229 * 255, 0.224 * 255, 0.225 * 255]
)
def __init__(self, batch_size, output_size, num_threads, device_id, images_directory):
super(HybridPipelineTrain, self).__init__(batch_size, num_threads, device_id, seed = 12)
self.input = ops.FileReader(file_root = images_directory, random_shuffle = True, initial_fill = 21)
self.decode = ops.ImageDecoder(device = "mixed", output_type = types.RGB)
self.cmn = ops.CropMirrorNormalize(
device="gpu",
dtype=types.FLOAT,
mean=[0.5, 0.5, 0.5],
std=[0.5, 0.5, 0.5],
output_layout="HWC"
)
self.coin = ops.random.CoinFlip(probability = 0.5)
self.flip = ops.Flip(device = "gpu")
self.rsz = ops.Resize(resize_x = output_size[0], resize_y = output_size[1], device = "gpu")
def __init__(self, batch_size, num_threads, device_id):
super(COCOPipeline, self).__init__(batch_size,
num_threads,
device_id,
exec_async=False,
exec_pipelined=False,
seed=15)
self.input = ops.COCOReader(file_root=file_root,
annotations_file=annotations_file,
shard_id=device_id,
num_shards=num_gpus,
ratio=True,
ltrb=True)
self.decode = ops.ImageDecoder(device="mixed", output_type=types.RGB)
self.flip = ops.Flip(device="gpu")
self.bbflip = ops.BbFlip(device="cpu", ltrb=True)
self.paste_pos = ops.Uniform(range=(0, 1))
self.paste_ratio = ops.Uniform(range=(1, 2))
self.coin = ops.CoinFlip(probability=0.5)
self.coin2 = ops.CoinFlip(probability=0.5)
self.paste = ops.Paste(device="gpu", fill_value=(32, 64, 128))
self.bbpaste = ops.BBoxPaste(device="cpu", ltrb=True)
self.prospective_crop = ops.RandomBBoxCrop(device="cpu",
aspect_ratio=[0.5, 2.0],
thresholds=[0.1, 0.3, 0.5],
scaling=[0.8, 1.0],
ltrb=True)
self.slice = ops.Slice(device="gpu")
# resize
self.resize = ops.Resize(device="gpu",
interp_type=types.INTERP_LINEAR,
resize_shorter=800,
max_size=1200)
self.shape = ops.Shapes(device="gpu")
# normalize and convert hwc to chw
self.cmnp = ops.CropMirrorNormalize(
device="gpu",
output_dtype=types.FLOAT,
output_layout=types.NCHW,
image_type=types.RGB,
mean=[0.485 * 255, 0.456 * 255, 0.406 * 255],
std=[0.229 * 255, 0.224 * 255, 0.225 * 255])
# padding axes=(0,1) -> hwc, axes=(1,2) -> chw
self.padding = ops.Pad(device="gpu",
fill_value=0,
axes=(1, 2),
shape=(800, 1200))
def __init__(self,
batch_size,
file_root,
annotations_file,
default_boxes,
seed,
device_id=0,
num_threads=4):
super(COCOPipeline, self).__init__(batch_size=batch_size,
device_id=device_id,
num_threads=num_threads,
seed=seed)
self.input = ops.COCOReader(file_root=file_root,
annotations_file=annotations_file,
ratio=True,
ltrb=True,
random_shuffle=True)
self.decode = ops.nvJPEGDecoder(device="mixed", output_type=types.RGB)
# Augumentation techniques
self.crop = ops.RandomBBoxCrop(device="cpu",
aspect_ratio=[0.5, 2.0],
thresholds=[0.1, 0.3, 0.5, 0.7, 0.9],
scaling=[0.8, 1.0],
ltrb=True)
self.slice = ops.Slice(device="gpu")
self.twist = ops.ColorTwist(device="gpu")
self.resize = ops.Resize(device="gpu", resize_x=300, resize_y=300)
self.normalize = ops.CropMirrorNormalize(
device="gpu",
crop=(300, 300),
mean=[0.485 * 255., 0.456 * 255., 0.406 * 255.],
std=[0.229 * 255., 0.224 * 255., 0.225 * 255.])
# Random variables
self.rng1 = ops.Uniform(range=[0.5, 1.5])
self.rng2 = ops.Uniform(range=[0.875, 1.125])
self.rng3 = ops.Uniform(range=[-0.5, 0.5])
self.flip = ops.Flip(device="gpu")
self.bbflip = ops.BbFlip(device="cpu", ltrb=True)
self.flip_coin = ops.CoinFlip(probability=0.5)
self.box_encoder = ops.BoxEncoder(device="cpu",
criteria=0.5,
anchors=default_boxes.as_ltrb_list())
def __init__(self, batch_size, output_size, num_threads, device_id, images_directory):
super(HybridPipelineTrain, self).__init__(batch_size, num_threads, device_id, seed = 12)
self.input = ops.FileReader(file_root = images_directory, random_shuffle = True, initial_fill = 21)
self.decode = ops.ImageDecoder(device = "mixed", output_type = types.RGB)
self.rotate = ops.Rotate(device = "gpu")
self.RRC = ops.RandomResizedCrop(size = output_size,
random_area = [0.4, 1.0],
random_aspect_ratio = [0.5, 1.5],
device="gpu"
)
self.cmn = ops.CropMirrorNormalize(
device="gpu",
dtype=types.FLOAT,
mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225],
output_layout="HWC"
)
self.rng = ops.random.Uniform(range = (-25.0, 25.0))
self.coin = ops.random.CoinFlip(probability = 0.5)
self.flip = ops.Flip(device = "gpu")
def __init__(self, img_file_list, num_shards, shard_id, batch_size,
num_threads, device_id):
super(Opt3Pipeline, self).__init__(batch_size,
num_threads,
device_id,
seed=12)
self.input_img = ops.FileReader(file_root="",
num_shards=num_shards,
shard_id=shard_id,
file_list=img_file_list,
random_shuffle=True,
initial_fill=21)
self.decode = ops.ImageDecoder(device="mixed", output_type=types.RGB)
self.rrc = ops.RandomResizedCrop(device="gpu",
size=(800, 800),
random_area=[0.8, 0.8])
self.flip_h = ops.Flip(device="gpu", vertical=0, horizontal=1)
self.rotate = ops.Rotate(device="gpu",
angle=30,
interp_type=types.INTERP_LINEAR,
fill_value=0)
self.resize = ops.Resize(device="gpu", resize_x=224, resize_y=224)
def __init__(self,
session,
dataset,
batch_size,
num_threads,
is_random_flip=True,
num_samples=1000000,
device_id=0,
preprocess=None,
fill_last_batch=True):
super(ExternalSourcePipeline, self).__init__(batch_size,
num_threads,
device_id,
seed=12)
self.session = session
self.num_samples = num_samples
self.dataset = dataset
self.is_random_flip = is_random_flip
self.preprocess = preprocess
if self.preprocess is not None:
crop = (preprocess['width'], preprocess['height'])
self.res = ops.Resize(resize_x=preprocess['width'],
resize_y=preprocess['height'])
self.flip = ops.Flip()
self.coin = ops.CoinFlip(probability=0.5)
self.coin2 = ops.CoinFlip(probability=0.5)
self.iterator = iter(
ExternalInputIterator3(self.session, self.dataset,
self.num_samples, batch_size,
fill_last_batch))
self.iterator.reset()
self.input = ops.ExternalSource()
self.input_label = ops.ExternalSource()
def __init__(self,
batch_size,
num_threads,
device_id,
external_data,
is_train=True):
super(ExternalSourcePipeline, self).__init__(batch_size,
num_threads,
device_id,
seed=12)
self.is_train = is_train
self.input = ops.ExternalSource()
self.angle_rng = ops.Uniform(range=(-10.0, 10.0))
self.rotate = ops.Rotate(device="gpu")
self.flip_rng = ops.CoinFlip(probability=0.5)
self.flip = ops.Flip(device='gpu')
self.slice = ops.Slice()
self.external_data = external_data
self.iterator = iter(self.external_data)