sc26-dl-tutorial/utils/data_loader_dali.py at main · NERSC/sc26-dl-tutorial · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
import h5py
import numpy as np
import torch
import torch.distributed as dist
import nvidia.dali.fn as fn
from nvidia.dali.plugin.pytorch import DALIGenericIterator, LastBatchPolicy
from nvidia.dali.pipeline import Pipeline

import utils.dali_es_helper as esh


def get_data_loader(
    params, data_path, distributed, train, spatial_shard=None
):
    dataloader = SynDaliDataLoader(
        params, data_path, train, spatial_shard=spatial_shard
    )

    if train:
        return dataloader, None, None
    else:
        return dataloader, None


class SynDaliDataLoader:
    def get_pipeline(self):
        pipeline = Pipeline(
            batch_size=self.batch_size,
            num_threads=2,
            device_id=self.device_index,
            py_num_workers=self.num_data_workers,
            py_start_method="spawn",
            seed=self.global_seed,
        )

        with pipeline:  # get input and target
            inp, tar = fn.external_source(
                source=esh.SynES(
                    self.data_path,
                    self.train,
                    self.batch_size,
                    self.dt,
                    self.img_size,
                    self.n_in_channels,
                    self.n_out_channels,
                    self.num_shards,
                    self.shard_id,
                    self.limit_nsamples,
                    spatial_shard=self.spatial_shard,
                    spatial_read_mode=self.spatial_read_mode,
                    enable_logging=False,
                    seed=self.global_seed,
                ),
                num_outputs=2,
                layout=["CHW", "CHW"],
                batch=False,
                parallel=True,
                prefetch_queue_depth=self.num_data_workers,
            )

            inp = inp.gpu()
            tar = tar.gpu()

            if self.normalize:
                inp = fn.normalize(
                    inp,
                    device="gpu",
                    axis_names="HW",
                    batch=False,
                    mean=self.in_bias,
                    stddev=self.in_scale,
                )

                tar = fn.normalize(
                    tar,
                    device="gpu",
                    axis_names="HW",
                    batch=False,
                    mean=self.out_bias,
                    stddev=self.out_scale,
                )

            pipeline.set_outputs(inp, tar)
        return pipeline

    def __init__(
        self, params, data_path, train, spatial_shard=None, seed=333
    ):
        self.global_seed = seed
        self.num_data_workers = params.num_data_workers
        self.device_index = torch.cuda.current_device()
        self.batch_size = int(params.local_batch_size)

        self.data_path = data_path
        self.train = train
        self.dt = params.dt
        self.n_in_channels = params.n_in_channels
        self.n_out_channels = params.n_out_channels
        self.img_size = params.img_size
        self.spatial_shard = spatial_shard
        self.spatial_read_mode = params.get("spatial_read_mode", "local")
        self.limit_nsamples = (
            params.limit_nsamples if train else params.limit_nsamples_val
        )

        with h5py.File(params.stats_path, "r") as stats:
            means = stats["mean"][:].astype(np.float32)[:, None, None]
            stds = stats["std"][:].astype(np.float32)[:, None, None]
        self.normalize = True
        self.in_bias = means[: self.n_in_channels]
        self.in_scale = stds[: self.n_in_channels]
        self.out_bias = means[: self.n_out_channels]
        self.out_scale = stds[: self.n_out_channels]

        if dist.is_initialized():
            self.num_shards = params.data_num_shards
            self.shard_id = params.data_shard_id
        else:
            self.num_shards = 1
            self.shard_id = 0

        extsource = esh.SynES(
            self.data_path,
            self.train,
            self.batch_size,
            self.dt,
            self.img_size,
            self.n_in_channels,
            self.n_out_channels,
            self.num_shards,
            self.shard_id,
            self.limit_nsamples,
            spatial_shard=self.spatial_shard,
            spatial_read_mode=self.spatial_read_mode,
            seed=self.global_seed,
        )
        self.num_batches = extsource.num_steps_per_epoch
        del extsource

        self.pipeline = self.get_pipeline()
        self.pipeline.start_py_workers()
        self.pipeline.build()

        self.iterator = DALIGenericIterator(
            [self.pipeline],
            ["inp", "tar"],
            auto_reset=True,
            last_batch_policy=LastBatchPolicy.DROP,
            prepare_first_batch=True,
        )

    def __len__(self):
        return self.num_batches

    def __iter__(self):
        for token in self.iterator:
            inp = token[0]["inp"]
            tar = token[0]["tar"]

            yield inp, tar