File size: 9,053 Bytes
ecd1674
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
# Copyright (c) 2024, NVIDIA Corporation & Affiliates. All rights reserved. 
# 
# This work is made available under the Nvidia Source Code License-NC. 
# To view a copy of this license, visit 
# https://github.com/NVlabs/QLIP/blob/main/LICENSE

# MIT License
# Based on https://github.com/zhaoyue-zephyrus/bsq-vit/blob/main/transcoder/models/quantizer/bsq.py

import torch
import torch.nn as nn
from einops import rearrange, reduce

_EPS = 1e-8


class DifferentiableEntropyFunction(torch.autograd.Function):
    @staticmethod
    def forward(ctx, zq, basis, K, eps):
        zb = (zq + 1) / 2
        zi = ((zb * basis).sum(-1)).to(torch.int64)
        cnt = torch.scatter_reduce(
            torch.zeros(2**K, device=zq.device, dtype=zq.dtype),
            0,
            zi.flatten(),
            torch.ones_like(zi.flatten()).to(zq.dtype),
            "sum",
        )
        prob = (cnt + eps) / (cnt + eps).sum()
        H = torch.special.entr(prob).sum()
        ctx.save_for_backward(zq, zi, prob)
        ctx.K = K
        return H

    @staticmethod
    def backward(ctx, grad_output):
        zq, zi, prob = ctx.saved_tensors
        grad_array = -grad_output * (torch.log(prob) + 1) / zi.numel() / ctx.K
        reord_grad = grad_array[zi.flatten()].reshape(zi.shape)
        grad_input = reord_grad.unsqueeze(-1) * zq
        return grad_input, None, None, None, None


def codebook_entropy(zq, basis, K, eps=1e-8):
    return DifferentiableEntropyFunction.apply(zq, basis, K, eps)


class BinarySphericalQuantizer(nn.Module):
    def __init__(
        self,
        embed_dim: int = 18,
        group_size: int = 9,
        soft_entropy: bool = True,
        beta: float = 0.0,  # commit loss
        gamma_0: float = 1.0,  # entropy loss (E[H(q)])
        gamma_1: float = 1.0,  # entropy loss (H[E[q]])
        input_format: str = "bchw",
        persample_entropy_compute: str = "group",
        l2_norm: bool = True,
        inv_temperature: float = 100.0,
    ):
        super().__init__()
        self.embed_dim = embed_dim
        self.group_size = group_size
        assert embed_dim % group_size == 0, "embed_dim must be divisible by group_size"
        self.soft_entropy = soft_entropy
        self.beta = beta
        self.gamma_0 = gamma_0
        self.gamma_1 = gamma_1
        assert input_format in ["bchw", "blc"]
        self.input_format = input_format
        assert persample_entropy_compute in [
            "group",
            "analytical",
        ], "persample_entropy_compute must be either 'group' or 'analytical'"
        self.persample_entropy_compute = persample_entropy_compute
        self.l2_norm = l2_norm
        self.inv_temperature = inv_temperature

        self.register_buffer("basis", 2 ** torch.arange(embed_dim - 1, -1, -1), persistent=False)
        self.register_buffer(
            "group_basis", 2 ** torch.arange(group_size - 1, -1, -1), persistent=False
        )

        group_codes = torch.arange(2**self.group_size)
        group_codebook = self.indexes_to_codes(group_codes).float()[:, -group_size:]
        self.register_buffer("group_codebook", group_codebook, persistent=False)

    def quantize(self, z):
        assert (
            z.shape[-1] == self.embed_dim
        ), f"Expected {self.embed_dim} dimensions, got {z.shape[-1]}"
        zhat = torch.where(z > 0, torch.ones_like(z), -torch.ones_like(z))
        return z + (zhat - z).detach()

    def forward(self, z):
        if self.input_format == "bchw":
            z = rearrange(z, "b c h w -> b h w c")
        zq = self.quantize(z)

        indices = self.codes_to_indexes(zq.detach())
        group_indices = self.codes_to_group_indexes(zq.detach())

        if not self.training:
            used_codes = torch.unique(indices, return_counts=False)
        else:
            used_codes = None

        if self.soft_entropy:
            persample_entropy, cb_entropy = self.soft_entropy_loss(z)
        else:
            persample_entropy, cb_entropy = self.hard_entropy_loss(z)
        entropy_penalty = self.gamma_0 * persample_entropy - self.gamma_1 * cb_entropy

        q_scale = 1.0 / (self.embed_dim**0.5) if self.l2_norm else 1.0
        zq = zq * q_scale
        commit_loss = self.beta * torch.mean(((zq.detach() - z) ** 2).sum(dim=-1))

        if self.input_format == "bchw":
            zq = rearrange(zq, "b h w c -> b c h w")

        return (
            zq,
            commit_loss + entropy_penalty / self.inv_temperature,
            {
                "H": cb_entropy,
                "used_codes": used_codes,
                "indices": indices,
                "group_indices": group_indices,
            },
        )

    def soft_entropy_loss(self, z):
        group_codebook = self.group_codebook / (self.embed_dim**0.5 if self.l2_norm else 1)
        divided_z = rearrange(z, "... (g c) -> ... g c", c=self.group_size)

        if self.persample_entropy_compute == "group":
            distance = -2 * torch.einsum("... g c, d c -> ... g d", divided_z, group_codebook)
            prob = (-distance * self.inv_temperature).softmax(dim=-1)
            persample_entropy = torch.special.entr(prob + _EPS).sum((-1, -2)).mean()
        else:
            p = torch.sigmoid(
                -4 * z / (self.embed_dim**0.5 if self.l2_norm else 1) * self.inv_temperature
            )
            prob = torch.stack([p, 1 - p], dim=-1)
            persample_entropy = torch.special.entr(prob + _EPS).sum((-1, -2)).mean()

        # macro average of the probability of each subgroup
        avg_prob = reduce(prob, "... g d -> g d", "mean")
        cb_entropy = torch.special.entr(avg_prob + _EPS).sum()

        return persample_entropy, cb_entropy

    def hard_entropy_loss(self, z):
        zb = ((z + 1) / 2).reshape(z.shape[0], -1, z.shape[-1]).to(torch.float32)
        prob_per_dim = zb.sum(1) / zb.shape[1]
        prob = torch.stack([prob_per_dim, 1 - prob_per_dim], dim=-1)
        persample_entropy = torch.special.entr(prob + _EPS).sum((-1, -2)).mean()
        cb_entropy = codebook_entropy(z, self.basis, self.embed_dim)

        return persample_entropy, cb_entropy

    def codes_to_indexes(self, zhat):
        """Converts a `code` to an index in the codebook.
        Args:
            zhat: A tensor of shape (B, ..., C) containing the codes. must be in {-1, 1}
        """
        assert (
            zhat.shape[-1] == self.embed_dim
        ), f"Expected {self.embed_dim} dimensions, got {zhat.shape[-1]}"
        return ((zhat.int() + 1) / 2 * self.basis).sum(axis=-1).to(torch.int64)

    def codes_to_group_indexes(self, zhat):
        """Converts a `code` to a list of indexes (in groups) in the codebook.
        Args:
            zhat: A tensor of shape (B, ..., C) containing the codes. must be in {-1, 1}
        """
        zhat_in_group = rearrange(zhat, "b ... (g c) -> b ... g c", c=self.group_size)
        return ((zhat_in_group.int() + 1) / 2 * self.group_basis).sum(axis=-1).to(torch.int64)

    def indexes_to_codes(self, indices):
        """Inverse of `codes_to_indexes`."""
        indices = indices.unsqueeze(-1)
        codes_non_centered = torch.remainder(torch.floor_divide(indices, self.basis), 2)
        return codes_non_centered * 2 - 1

    def group_indexes_to_codes(self, group_indices):
        """Inverse of `codes_to_group_indexes`."""
        group_indices = group_indices.unsqueeze(-1)
        codes_non_centered = torch.remainder(torch.floor_divide(group_indices, self.group_basis), 2)
        codes_non_centered = rearrange(codes_non_centered, "b ... g c -> b ... (g c)")
        return codes_non_centered * 2 - 1

    def get_group_codebook_entry(self, group_indices, one_hot=False):
        """
        Args:
            group_indices: A tensor of shape (B, L, G, C) containing the group indices.
        """
        if one_hot:
            z_q = group_indices @ self.group_codebook
        else:
            z_q = self.group_indexes_to_codes(group_indices)
        q_scale = 1.0 / (self.embed_dim**0.5) if self.l2_norm else 1.0
        z_q = z_q * q_scale
        if self.input_format == "bchw":
            h, w = int(z_q.shape[1] ** 0.5)
            assert h * w == z_q.shape[1], "Invalid sequence length"
            z_q = rearrange(z_q, "b (h w) c -> b c h w", h=h)
        return z_q

    def get_codebook_entry(self, indices, one_hot=False):
        """
        Args:
            group_indices: A tensor of shape (B, L, C) containing the indices.
        """
        if one_hot:
            assert self.embed_dim == self.group_size, "one_hot is only supported for group_size == embed_dim"
            z_q = indices @ self.group_codebook
        else:
            z_q = self.indexes_to_codes(indices)
        q_scale = 1.0 / (self.embed_dim**0.5) if self.l2_norm else 1.0
        z_q = z_q * q_scale
        if self.input_format == "bchw":
            h, w = int(z_q.shape[1] ** 0.5)
            assert h * w == z_q.shape[1], "Invalid sequence length"
            z_q = rearrange(z_q, "b (h w) c -> b c h w", h=h)
        return z_q