File size: 8,650 Bytes
5085882
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
# Copyright (c) Meta Platforms, Inc. and affiliates.
# All rights reserved.

# This source code is licensed under the license found in the
# LICENSE file in the root directory of this source tree.
# --------------------------------------------------------
# Position embedding utils
# --------------------------------------------------------

import numpy as np

import torch

# --------------------------------------------------------
# 2D sine-cosine position embedding
# References:
# Transformer: https://github.com/tensorflow/models/blob/master/official/nlp/transformer/model_utils.py
# MoCo v3: https://github.com/facebookresearch/moco-v3
# --------------------------------------------------------
def get_2d_sincos_pos_embed(embed_dim, grid_size, cls_token=False):
    """
    grid_size: int of the grid height and width
    return:
    pos_embed: [grid_size*grid_size, embed_dim] or [1+grid_size*grid_size, embed_dim] (w/ or w/o cls_token)
    """
    grid_h = np.arange(grid_size, dtype=np.float32)
    grid_w = np.arange(grid_size, dtype=np.float32)
    grid = np.meshgrid(grid_w, grid_h)  # here w goes first
    grid = np.stack(grid, axis=0)

    grid = grid.reshape([2, 1, grid_size, grid_size])
    pos_embed = get_2d_sincos_pos_embed_from_grid(embed_dim, grid)
    if cls_token:
        pos_embed = np.concatenate([np.zeros([1, embed_dim]), pos_embed], axis=0)
    return pos_embed


def get_2d_sincos_pos_embed_flexible(embed_dim, grid_size, cls_token=False):
    """
    grid_size: int of the grid height and width
    return:
    pos_embed: [grid_size*grid_size, embed_dim] or [1+grid_size*grid_size, embed_dim] (w/ or w/o cls_token)
    """
    grid_h = np.arange(grid_size[0], dtype=np.float32)
    grid_w = np.arange(grid_size[1], dtype=np.float32)
    grid = np.meshgrid(grid_w, grid_h)  # here w goes first
    grid = np.stack(grid, axis=0)

    grid = grid.reshape([2, 1, grid_size[0], grid_size[1]])
    pos_embed = get_2d_sincos_pos_embed_from_grid(embed_dim, grid)
    if cls_token:
        pos_embed = np.concatenate([np.zeros([1, embed_dim]), pos_embed], axis=0)
    return pos_embed


def get_2d_sincos_pos_embed_from_grid(embed_dim, grid):
    assert embed_dim % 2 == 0

    # use half of dimensions to encode grid_h
    emb_h = get_1d_sincos_pos_embed_from_grid(embed_dim // 2, grid[0])  # (H*W, D/2)
    emb_w = get_1d_sincos_pos_embed_from_grid(embed_dim // 2, grid[1])  # (H*W, D/2)

    emb = np.concatenate([emb_h, emb_w], axis=1)  # (H*W, D)
    return emb


def get_1d_sincos_pos_embed_from_grid(embed_dim, pos):
    """
    embed_dim: output dimension for each position
    pos: a list of positions to be encoded: size (M,)
    out: (M, D)
    """
    assert embed_dim % 2 == 0
    # omega = np.arange(embed_dim // 2, dtype=np.float)
    omega = np.arange(embed_dim // 2, dtype=float)
    omega /= embed_dim / 2.0
    omega = 1.0 / 10000**omega  # (D/2,)

    pos = pos.reshape(-1)  # (M,)
    out = np.einsum("m,d->md", pos, omega)  # (M, D/2), outer product

    emb_sin = np.sin(out)  # (M, D/2)
    emb_cos = np.cos(out)  # (M, D/2)

    emb = np.concatenate([emb_sin, emb_cos], axis=1)  # (M, D)
    return emb


# --------------------------------------------------------
# Interpolate position embeddings for high-resolution
# References:
# DeiT: https://github.com/facebookresearch/deit
# --------------------------------------------------------
def interpolate_pos_embed(model, checkpoint_model):
    if "pos_embed" in checkpoint_model:
        pos_embed_checkpoint = checkpoint_model["pos_embed"]
        embedding_size = pos_embed_checkpoint.shape[-1]
        num_patches = model.patch_embed.num_patches
        num_extra_tokens = model.pos_embed.shape[-2] - num_patches
        # height (== width) for the checkpoint position embedding
        orig_size = int((pos_embed_checkpoint.shape[-2] - num_extra_tokens) ** 0.5)
        # height (== width) for the new position embedding
        new_size = int(num_patches**0.5)
        # class_token and dist_token are kept unchanged
        if orig_size != new_size:
            print(
                "Position interpolate from %dx%d to %dx%d"
                % (orig_size, orig_size, new_size, new_size)
            )
            extra_tokens = pos_embed_checkpoint[:, :num_extra_tokens]
            # only the position tokens are interpolated
            pos_tokens = pos_embed_checkpoint[:, num_extra_tokens:]
            pos_tokens = pos_tokens.reshape(
                -1, orig_size, orig_size, embedding_size
            ).permute(0, 3, 1, 2)
            pos_tokens = torch.nn.functional.interpolate(
                pos_tokens,
                size=(new_size, new_size),
                mode="bicubic",
                align_corners=False,
            )
            pos_tokens = pos_tokens.permute(0, 2, 3, 1).flatten(1, 2)
            new_pos_embed = torch.cat((extra_tokens, pos_tokens), dim=1)
            checkpoint_model["pos_embed"] = new_pos_embed


def interpolate_pos_embed_img2audio(model, checkpoint_model, orig_size, new_size):
    if "pos_embed" in checkpoint_model:
        pos_embed_checkpoint = checkpoint_model["pos_embed"]
        embedding_size = pos_embed_checkpoint.shape[-1]
        num_patches = model.patch_embed.num_patches
        num_extra_tokens = model.pos_embed.shape[-2] - num_patches
        # height (== width) for the checkpoint position embedding
        # orig_size = int((pos_embed_checkpoint.shape[-2] - num_extra_tokens) ** 0.5)
        # height (== width) for the new position embedding
        # new_size = int(num_patches ** 0.5)
        # class_token and dist_token are kept unchanged
        if orig_size != new_size:
            print(
                "Position interpolate from %dx%d to %dx%d"
                % (orig_size[0], orig_size[1], new_size[0], new_size[1])
            )
            extra_tokens = pos_embed_checkpoint[:, :num_extra_tokens]
            # only the position tokens are interpolated
            pos_tokens = pos_embed_checkpoint[:, num_extra_tokens:]
            pos_tokens = pos_tokens.reshape(
                -1, orig_size[0], orig_size[1], embedding_size
            ).permute(0, 3, 1, 2)
            pos_tokens = torch.nn.functional.interpolate(
                pos_tokens,
                size=(new_size[0], new_size[1]),
                mode="bicubic",
                align_corners=False,
            )
            pos_tokens = pos_tokens.permute(0, 2, 3, 1).flatten(1, 2)
            new_pos_embed = torch.cat((extra_tokens, pos_tokens), dim=1)
            checkpoint_model["pos_embed"] = new_pos_embed


def interpolate_pos_embed_audio(model, checkpoint_model, orig_size, new_size):
    if "pos_embed" in checkpoint_model:
        pos_embed_checkpoint = checkpoint_model["pos_embed"]
        embedding_size = pos_embed_checkpoint.shape[-1]
        num_patches = model.patch_embed.num_patches
        num_extra_tokens = model.pos_embed.shape[-2] - num_patches
        if orig_size != new_size:
            print(
                "Position interpolate from %dx%d to %dx%d"
                % (orig_size[0], orig_size[1], new_size[0], new_size[1])
            )
            # extra_tokens = pos_embed_checkpoint[:, :num_extra_tokens]
            # only the position tokens are interpolated
            cls_token = pos_embed_checkpoint[:, 0, :].unsqueeze(1)
            pos_tokens = pos_embed_checkpoint[:, 1:, :]  # remove
            pos_tokens = pos_tokens.reshape(
                -1, orig_size[0], orig_size[1], embedding_size
            )  # .permute(0, 3, 1, 2)
            # pos_tokens = torch.nn.functional.interpolate(
            #    pos_tokens, size=(new_size[0], new_size[1]), mode='bicubic', align_corners=False)

            # pos_tokens = pos_tokens.permute(0, 2, 3, 1).flatten(1, 2)
            pos_tokens = pos_tokens[:, :, : new_size[1], :]  # assume only time diff
            pos_tokens = pos_tokens.flatten(1, 2)
            new_pos_embed = torch.cat((cls_token, pos_tokens), dim=1)
            checkpoint_model["pos_embed"] = new_pos_embed


def interpolate_patch_embed_audio(
    model,
    checkpoint_model,
    orig_channel,
    new_channel=1,
    kernel_size=(16, 16),
    stride=(16, 16),
    padding=(0, 0),
):
    if orig_channel != new_channel:
        if "patch_embed.proj.weight" in checkpoint_model:
            # aggregate 3 channels in rgb ckpt to 1 channel for audio
            new_proj_weight = torch.nn.Parameter(
                torch.sum(checkpoint_model["patch_embed.proj.weight"], dim=1).unsqueeze(
                    1
                )
            )
            checkpoint_model["patch_embed.proj.weight"] = new_proj_weight