File size: 5,264 Bytes
0034848
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
"""
Copyright (c) Microsoft Corporation.
Licensed under the MIT license.

This file contains primitives for multi-gpu communication.
This is useful when doing distributed training.
"""

import pickle
import time

import torch
import torch.distributed as dist

from comfy.model_management import get_torch_device
device = get_torch_device()


def get_world_size():
    if not dist.is_available():
        return 1
    if not dist.is_initialized():
        return 1
    return dist.get_world_size()


def get_rank():
    if not dist.is_available():
        return 0
    if not dist.is_initialized():
        return 0
    return dist.get_rank()


def is_main_process():
    return get_rank() == 0


def synchronize():
    """
    Helper function to synchronize (barrier) among all processes when
    using distributed training
    """
    if not dist.is_available():
        return
    if not dist.is_initialized():
        return
    world_size = dist.get_world_size()
    if world_size == 1:
        return
    dist.barrier()


def gather_on_master(data):
    """Same as all_gather, but gathers data on master process only, using CPU.
    Thus, this does not work with NCCL backend unless they add CPU support.

    The memory consumption of this function is ~ 3x of data size. While in
    principal, it should be ~2x, it's not easy to force Python to release
    memory immediately and thus, peak memory usage could be up to 3x.
    """
    world_size = get_world_size()
    if world_size == 1:
        return [data]

    # serialized to a Tensor
    buffer = pickle.dumps(data)
    # trying to optimize memory, but in fact, it's not guaranteed to be released
    del data
    storage = torch.ByteStorage.from_buffer(buffer)
    del buffer
    tensor = torch.ByteTensor(storage)

    # obtain Tensor size of each rank
    local_size = torch.LongTensor([tensor.numel()])
    size_list = [torch.LongTensor([0]) for _ in range(world_size)]
    dist.all_gather(size_list, local_size)
    size_list = [int(size.item()) for size in size_list]
    max_size = max(size_list)

    if local_size != max_size:
        padding = torch.ByteTensor(size=(max_size - local_size,))
        tensor = torch.cat((tensor, padding), dim=0)
        del padding

    if is_main_process():
        tensor_list = []
        for _ in size_list:
            tensor_list.append(torch.ByteTensor(size=(max_size,)))
        dist.gather(tensor, gather_list=tensor_list, dst=0)
        del tensor
    else:
        dist.gather(tensor, gather_list=[], dst=0)
        del tensor
        return

    data_list = []
    for tensor in tensor_list:
        buffer = tensor.cpu().numpy().tobytes()
        del tensor
        data_list.append(pickle.loads(buffer))
        del buffer

    return data_list


def all_gather(data):
    """
    Run all_gather on arbitrary picklable data (not necessarily tensors)
    Args:
        data: any picklable object
    Returns:
        list[data]: list of data gathered from each rank
    """
    world_size = get_world_size()
    if world_size == 1:
        return [data]

    # serialized to a Tensor
    buffer = pickle.dumps(data)
    storage = torch.ByteStorage.from_buffer(buffer)
    tensor = torch.ByteTensor(storage).to(device)

    # obtain Tensor size of each rank
    local_size = torch.LongTensor([tensor.numel()]).to(device)
    size_list = [torch.LongTensor([0]).to(device) for _ in range(world_size)]
    dist.all_gather(size_list, local_size)
    size_list = [int(size.item()) for size in size_list]
    max_size = max(size_list)

    # receiving Tensor from all ranks
    # we pad the tensor because torch all_gather does not support
    # gathering tensors of different shapes
    tensor_list = []
    for _ in size_list:
        tensor_list.append(torch.ByteTensor(size=(max_size,)).to(device))
    if local_size != max_size:
        padding = torch.ByteTensor(size=(max_size - local_size,)).to(device)
        tensor = torch.cat((tensor, padding), dim=0)
    dist.all_gather(tensor_list, tensor)

    data_list = []
    for size, tensor in zip(size_list, tensor_list):
        buffer = tensor.cpu().numpy().tobytes()[:size]
        data_list.append(pickle.loads(buffer))

    return data_list


def reduce_dict(input_dict, average=True):
    """
    Args:
        input_dict (dict): all the values will be reduced
        average (bool): whether to do average or sum
    Reduce the values in the dictionary from all processes so that process with rank
    0 has the averaged results. Returns a dict with the same fields as
    input_dict, after reduction.
    """
    world_size = get_world_size()
    if world_size < 2:
        return input_dict
    with torch.no_grad():
        names = []
        values = []
        # sort the keys so that they are consistent across processes
        for k in sorted(input_dict.keys()):
            names.append(k)
            values.append(input_dict[k])
        values = torch.stack(values, dim=0)
        dist.reduce(values, dst=0)
        if dist.get_rank() == 0 and average:
            # only main process gets accumulated, so only divide by
            # world_size in this case
            values /= world_size
        reduced_dict = {k: v for k, v in zip(names, values)}
    return reduced_dict