final upload
Browse files- models/diffloss.py +248 -0
- models/mar.py +355 -0
- models/vae.py +67 -0
- taming/modules/autoencoder/lpips/vgg.pth +3 -0
- util/crop.py +23 -0
- util/download.py +62 -0
- util/loader.py +56 -0
- util/lr_sched.py +21 -0
- util/misc.py +340 -0
models/diffloss.py
ADDED
@@ -0,0 +1,248 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import torch
|
2 |
+
import torch.nn as nn
|
3 |
+
from torch.utils.checkpoint import checkpoint
|
4 |
+
import math
|
5 |
+
|
6 |
+
from diffusion import create_diffusion
|
7 |
+
|
8 |
+
|
9 |
+
class DiffLoss(nn.Module):
|
10 |
+
"""Diffusion Loss"""
|
11 |
+
def __init__(self, target_channels, z_channels, depth, width, num_sampling_steps, grad_checkpointing=False):
|
12 |
+
super(DiffLoss, self).__init__()
|
13 |
+
self.in_channels = target_channels
|
14 |
+
self.net = SimpleMLPAdaLN(
|
15 |
+
in_channels=target_channels,
|
16 |
+
model_channels=width,
|
17 |
+
out_channels=target_channels * 2, # for vlb loss
|
18 |
+
z_channels=z_channels,
|
19 |
+
num_res_blocks=depth,
|
20 |
+
grad_checkpointing=grad_checkpointing
|
21 |
+
)
|
22 |
+
|
23 |
+
self.train_diffusion = create_diffusion(timestep_respacing="", noise_schedule="cosine")
|
24 |
+
self.gen_diffusion = create_diffusion(timestep_respacing=num_sampling_steps, noise_schedule="cosine")
|
25 |
+
|
26 |
+
def forward(self, target, z, mask=None):
|
27 |
+
t = torch.randint(0, self.train_diffusion.num_timesteps, (target.shape[0],), device=target.device)
|
28 |
+
model_kwargs = dict(c=z)
|
29 |
+
loss_dict = self.train_diffusion.training_losses(self.net, target, t, model_kwargs)
|
30 |
+
loss = loss_dict["loss"]
|
31 |
+
if mask is not None:
|
32 |
+
loss = (loss * mask).sum() / mask.sum()
|
33 |
+
return loss.mean()
|
34 |
+
|
35 |
+
def sample(self, z, temperature=1.0, cfg=1.0):
|
36 |
+
# diffusion loss sampling
|
37 |
+
if not cfg == 1.0:
|
38 |
+
noise = torch.randn(z.shape[0] // 2, self.in_channels).cuda()
|
39 |
+
noise = torch.cat([noise, noise], dim=0)
|
40 |
+
model_kwargs = dict(c=z, cfg_scale=cfg)
|
41 |
+
sample_fn = self.net.forward_with_cfg
|
42 |
+
else:
|
43 |
+
noise = torch.randn(z.shape[0], self.in_channels).cuda()
|
44 |
+
model_kwargs = dict(c=z)
|
45 |
+
sample_fn = self.net.forward
|
46 |
+
|
47 |
+
sampled_token_latent = self.gen_diffusion.p_sample_loop(
|
48 |
+
sample_fn, noise.shape, noise, clip_denoised=False, model_kwargs=model_kwargs, progress=False,
|
49 |
+
temperature=temperature
|
50 |
+
)
|
51 |
+
|
52 |
+
return sampled_token_latent
|
53 |
+
|
54 |
+
|
55 |
+
def modulate(x, shift, scale):
|
56 |
+
return x * (1 + scale) + shift
|
57 |
+
|
58 |
+
|
59 |
+
class TimestepEmbedder(nn.Module):
|
60 |
+
"""
|
61 |
+
Embeds scalar timesteps into vector representations.
|
62 |
+
"""
|
63 |
+
def __init__(self, hidden_size, frequency_embedding_size=256):
|
64 |
+
super().__init__()
|
65 |
+
self.mlp = nn.Sequential(
|
66 |
+
nn.Linear(frequency_embedding_size, hidden_size, bias=True),
|
67 |
+
nn.SiLU(),
|
68 |
+
nn.Linear(hidden_size, hidden_size, bias=True),
|
69 |
+
)
|
70 |
+
self.frequency_embedding_size = frequency_embedding_size
|
71 |
+
|
72 |
+
@staticmethod
|
73 |
+
def timestep_embedding(t, dim, max_period=10000):
|
74 |
+
"""
|
75 |
+
Create sinusoidal timestep embeddings.
|
76 |
+
:param t: a 1-D Tensor of N indices, one per batch element.
|
77 |
+
These may be fractional.
|
78 |
+
:param dim: the dimension of the output.
|
79 |
+
:param max_period: controls the minimum frequency of the embeddings.
|
80 |
+
:return: an (N, D) Tensor of positional embeddings.
|
81 |
+
"""
|
82 |
+
# https://github.com/openai/glide-text2im/blob/main/glide_text2im/nn.py
|
83 |
+
half = dim // 2
|
84 |
+
freqs = torch.exp(
|
85 |
+
-math.log(max_period) * torch.arange(start=0, end=half, dtype=torch.float32) / half
|
86 |
+
).to(device=t.device)
|
87 |
+
args = t[:, None].float() * freqs[None]
|
88 |
+
embedding = torch.cat([torch.cos(args), torch.sin(args)], dim=-1)
|
89 |
+
if dim % 2:
|
90 |
+
embedding = torch.cat([embedding, torch.zeros_like(embedding[:, :1])], dim=-1)
|
91 |
+
return embedding
|
92 |
+
|
93 |
+
def forward(self, t):
|
94 |
+
t_freq = self.timestep_embedding(t, self.frequency_embedding_size)
|
95 |
+
t_emb = self.mlp(t_freq)
|
96 |
+
return t_emb
|
97 |
+
|
98 |
+
|
99 |
+
class ResBlock(nn.Module):
|
100 |
+
"""
|
101 |
+
A residual block that can optionally change the number of channels.
|
102 |
+
:param channels: the number of input channels.
|
103 |
+
"""
|
104 |
+
|
105 |
+
def __init__(
|
106 |
+
self,
|
107 |
+
channels
|
108 |
+
):
|
109 |
+
super().__init__()
|
110 |
+
self.channels = channels
|
111 |
+
|
112 |
+
self.in_ln = nn.LayerNorm(channels, eps=1e-6)
|
113 |
+
self.mlp = nn.Sequential(
|
114 |
+
nn.Linear(channels, channels, bias=True),
|
115 |
+
nn.SiLU(),
|
116 |
+
nn.Linear(channels, channels, bias=True),
|
117 |
+
)
|
118 |
+
|
119 |
+
self.adaLN_modulation = nn.Sequential(
|
120 |
+
nn.SiLU(),
|
121 |
+
nn.Linear(channels, 3 * channels, bias=True)
|
122 |
+
)
|
123 |
+
|
124 |
+
def forward(self, x, y):
|
125 |
+
shift_mlp, scale_mlp, gate_mlp = self.adaLN_modulation(y).chunk(3, dim=-1)
|
126 |
+
h = modulate(self.in_ln(x), shift_mlp, scale_mlp)
|
127 |
+
h = self.mlp(h)
|
128 |
+
return x + gate_mlp * h
|
129 |
+
|
130 |
+
|
131 |
+
class FinalLayer(nn.Module):
|
132 |
+
"""
|
133 |
+
The final layer adopted from DiT.
|
134 |
+
"""
|
135 |
+
def __init__(self, model_channels, out_channels):
|
136 |
+
super().__init__()
|
137 |
+
self.norm_final = nn.LayerNorm(model_channels, elementwise_affine=False, eps=1e-6)
|
138 |
+
self.linear = nn.Linear(model_channels, out_channels, bias=True)
|
139 |
+
self.adaLN_modulation = nn.Sequential(
|
140 |
+
nn.SiLU(),
|
141 |
+
nn.Linear(model_channels, 2 * model_channels, bias=True)
|
142 |
+
)
|
143 |
+
|
144 |
+
def forward(self, x, c):
|
145 |
+
shift, scale = self.adaLN_modulation(c).chunk(2, dim=-1)
|
146 |
+
x = modulate(self.norm_final(x), shift, scale)
|
147 |
+
x = self.linear(x)
|
148 |
+
return x
|
149 |
+
|
150 |
+
|
151 |
+
class SimpleMLPAdaLN(nn.Module):
|
152 |
+
"""
|
153 |
+
The MLP for Diffusion Loss.
|
154 |
+
:param in_channels: channels in the input Tensor.
|
155 |
+
:param model_channels: base channel count for the model.
|
156 |
+
:param out_channels: channels in the output Tensor.
|
157 |
+
:param z_channels: channels in the condition.
|
158 |
+
:param num_res_blocks: number of residual blocks per downsample.
|
159 |
+
"""
|
160 |
+
|
161 |
+
def __init__(
|
162 |
+
self,
|
163 |
+
in_channels,
|
164 |
+
model_channels,
|
165 |
+
out_channels,
|
166 |
+
z_channels,
|
167 |
+
num_res_blocks,
|
168 |
+
grad_checkpointing=False
|
169 |
+
):
|
170 |
+
super().__init__()
|
171 |
+
|
172 |
+
self.in_channels = in_channels
|
173 |
+
self.model_channels = model_channels
|
174 |
+
self.out_channels = out_channels
|
175 |
+
self.num_res_blocks = num_res_blocks
|
176 |
+
self.grad_checkpointing = grad_checkpointing
|
177 |
+
|
178 |
+
self.time_embed = TimestepEmbedder(model_channels)
|
179 |
+
self.cond_embed = nn.Linear(z_channels, model_channels)
|
180 |
+
|
181 |
+
self.input_proj = nn.Linear(in_channels, model_channels)
|
182 |
+
|
183 |
+
res_blocks = []
|
184 |
+
for i in range(num_res_blocks):
|
185 |
+
res_blocks.append(ResBlock(
|
186 |
+
model_channels,
|
187 |
+
))
|
188 |
+
|
189 |
+
self.res_blocks = nn.ModuleList(res_blocks)
|
190 |
+
self.final_layer = FinalLayer(model_channels, out_channels)
|
191 |
+
|
192 |
+
self.initialize_weights()
|
193 |
+
|
194 |
+
def initialize_weights(self):
|
195 |
+
def _basic_init(module):
|
196 |
+
if isinstance(module, nn.Linear):
|
197 |
+
torch.nn.init.xavier_uniform_(module.weight)
|
198 |
+
if module.bias is not None:
|
199 |
+
nn.init.constant_(module.bias, 0)
|
200 |
+
self.apply(_basic_init)
|
201 |
+
|
202 |
+
# Initialize timestep embedding MLP
|
203 |
+
nn.init.normal_(self.time_embed.mlp[0].weight, std=0.02)
|
204 |
+
nn.init.normal_(self.time_embed.mlp[2].weight, std=0.02)
|
205 |
+
|
206 |
+
# Zero-out adaLN modulation layers
|
207 |
+
for block in self.res_blocks:
|
208 |
+
nn.init.constant_(block.adaLN_modulation[-1].weight, 0)
|
209 |
+
nn.init.constant_(block.adaLN_modulation[-1].bias, 0)
|
210 |
+
|
211 |
+
# Zero-out output layers
|
212 |
+
nn.init.constant_(self.final_layer.adaLN_modulation[-1].weight, 0)
|
213 |
+
nn.init.constant_(self.final_layer.adaLN_modulation[-1].bias, 0)
|
214 |
+
nn.init.constant_(self.final_layer.linear.weight, 0)
|
215 |
+
nn.init.constant_(self.final_layer.linear.bias, 0)
|
216 |
+
|
217 |
+
def forward(self, x, t, c):
|
218 |
+
"""
|
219 |
+
Apply the model to an input batch.
|
220 |
+
:param x: an [N x C] Tensor of inputs.
|
221 |
+
:param t: a 1-D batch of timesteps.
|
222 |
+
:param c: conditioning from AR transformer.
|
223 |
+
:return: an [N x C] Tensor of outputs.
|
224 |
+
"""
|
225 |
+
x = self.input_proj(x)
|
226 |
+
t = self.time_embed(t)
|
227 |
+
c = self.cond_embed(c)
|
228 |
+
|
229 |
+
y = t + c
|
230 |
+
|
231 |
+
if self.grad_checkpointing and not torch.jit.is_scripting():
|
232 |
+
for block in self.res_blocks:
|
233 |
+
x = checkpoint(block, x, y)
|
234 |
+
else:
|
235 |
+
for block in self.res_blocks:
|
236 |
+
x = block(x, y)
|
237 |
+
|
238 |
+
return self.final_layer(x, y)
|
239 |
+
|
240 |
+
def forward_with_cfg(self, x, t, c, cfg_scale):
|
241 |
+
half = x[: len(x) // 2]
|
242 |
+
combined = torch.cat([half, half], dim=0)
|
243 |
+
model_out = self.forward(combined, t, c)
|
244 |
+
eps, rest = model_out[:, :self.in_channels], model_out[:, self.in_channels:]
|
245 |
+
cond_eps, uncond_eps = torch.split(eps, len(eps) // 2, dim=0)
|
246 |
+
half_eps = uncond_eps + cfg_scale * (cond_eps - uncond_eps)
|
247 |
+
eps = torch.cat([half_eps, half_eps], dim=0)
|
248 |
+
return torch.cat([eps, rest], dim=1)
|
models/mar.py
ADDED
@@ -0,0 +1,355 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from functools import partial
|
2 |
+
|
3 |
+
import numpy as np
|
4 |
+
from tqdm import tqdm
|
5 |
+
import scipy.stats as stats
|
6 |
+
import math
|
7 |
+
import torch
|
8 |
+
import torch.nn as nn
|
9 |
+
from torch.utils.checkpoint import checkpoint
|
10 |
+
|
11 |
+
from timm.models.vision_transformer import Block
|
12 |
+
|
13 |
+
from models.diffloss import DiffLoss
|
14 |
+
|
15 |
+
|
16 |
+
def mask_by_order(mask_len, order, bsz, seq_len):
|
17 |
+
masking = torch.zeros(bsz, seq_len).cuda()
|
18 |
+
masking = torch.scatter(masking, dim=-1, index=order[:, :mask_len.long()], src=torch.ones(bsz, seq_len).cuda()).bool()
|
19 |
+
return masking
|
20 |
+
|
21 |
+
|
22 |
+
class MAR(nn.Module):
|
23 |
+
""" Masked Autoencoder with VisionTransformer backbone
|
24 |
+
"""
|
25 |
+
def __init__(self, img_size=256, vae_stride=16, patch_size=1,
|
26 |
+
encoder_embed_dim=1024, encoder_depth=16, encoder_num_heads=16,
|
27 |
+
decoder_embed_dim=1024, decoder_depth=16, decoder_num_heads=16,
|
28 |
+
mlp_ratio=4., norm_layer=nn.LayerNorm,
|
29 |
+
vae_embed_dim=16,
|
30 |
+
mask_ratio_min=0.7,
|
31 |
+
label_drop_prob=0.1,
|
32 |
+
attn_dropout=0.1,
|
33 |
+
proj_dropout=0.1,
|
34 |
+
buffer_size=64,
|
35 |
+
diffloss_d=3,
|
36 |
+
diffloss_w=1024,
|
37 |
+
num_sampling_steps='100',
|
38 |
+
diffusion_batch_mul=4,
|
39 |
+
grad_checkpointing=False,
|
40 |
+
):
|
41 |
+
super().__init__()
|
42 |
+
|
43 |
+
# --------------------------------------------------------------------------
|
44 |
+
# VAE and patchify specifics
|
45 |
+
self.vae_embed_dim = vae_embed_dim
|
46 |
+
|
47 |
+
self.img_size = img_size
|
48 |
+
self.vae_stride = vae_stride
|
49 |
+
self.patch_size = patch_size
|
50 |
+
self.seq_h = self.seq_w = img_size // vae_stride // patch_size
|
51 |
+
self.seq_len = self.seq_h * self.seq_w
|
52 |
+
self.token_embed_dim = vae_embed_dim * patch_size**2
|
53 |
+
self.grad_checkpointing = grad_checkpointing
|
54 |
+
|
55 |
+
# --------------------------------------------------------------------------
|
56 |
+
# image drop
|
57 |
+
self.label_drop_prob = label_drop_prob
|
58 |
+
# Fake class embedding for CFG's unconditional generation
|
59 |
+
# self.fake_latent = nn.Parameter(torch.zeros(1, encoder_embed_dim))
|
60 |
+
|
61 |
+
# --------------------------------------------------------------------------
|
62 |
+
# MAR variant masking ratio, a left-half truncated Gaussian centered at 100% masking ratio with std 0.25
|
63 |
+
self.mask_ratio_generator = stats.truncnorm((mask_ratio_min - 1.0) / 0.25, 0, loc=1.0, scale=0.25)
|
64 |
+
|
65 |
+
# --------------------------------------------------------------------------
|
66 |
+
# MAR encoder specifics
|
67 |
+
self.z_proj1 = nn.Linear(self.token_embed_dim, encoder_embed_dim, bias=True)
|
68 |
+
self.z_proj2 = nn.Linear(self.token_embed_dim, encoder_embed_dim, bias=True)
|
69 |
+
self.z_proj_ln = nn.LayerNorm(encoder_embed_dim, eps=1e-6)
|
70 |
+
self.buffer_size = buffer_size
|
71 |
+
self.encoder_pos_embed_learned = nn.Parameter(torch.zeros(1, 2 * self.seq_len, encoder_embed_dim))
|
72 |
+
|
73 |
+
self.encoder_blocks = nn.ModuleList([
|
74 |
+
Block(encoder_embed_dim, encoder_num_heads, mlp_ratio, qkv_bias=True, norm_layer=norm_layer,
|
75 |
+
proj_drop=proj_dropout, attn_drop=attn_dropout) for _ in range(encoder_depth)])
|
76 |
+
self.encoder_norm = norm_layer(encoder_embed_dim)
|
77 |
+
|
78 |
+
# --------------------------------------------------------------------------
|
79 |
+
# MAR decoder specifics
|
80 |
+
self.decoder_embed = nn.Linear(encoder_embed_dim, decoder_embed_dim, bias=True)
|
81 |
+
self.mask_token = nn.Parameter(torch.zeros(1, 1, decoder_embed_dim))
|
82 |
+
self.decoder_pos_embed_learned = nn.Parameter(torch.zeros(1, 2 * self.seq_len, decoder_embed_dim))
|
83 |
+
|
84 |
+
self.decoder_blocks = nn.ModuleList([
|
85 |
+
Block(decoder_embed_dim, decoder_num_heads, mlp_ratio, qkv_bias=True, norm_layer=norm_layer,
|
86 |
+
proj_drop=proj_dropout, attn_drop=attn_dropout) for _ in range(decoder_depth)])
|
87 |
+
|
88 |
+
self.decoder_norm = norm_layer(decoder_embed_dim)
|
89 |
+
self.diffusion_pos_embed_learned = nn.Parameter(torch.zeros(1, 2*self.seq_len, decoder_embed_dim))
|
90 |
+
|
91 |
+
self.initialize_weights()
|
92 |
+
|
93 |
+
# --------------------------------------------------------------------------
|
94 |
+
# Diffusion Loss
|
95 |
+
self.diffloss = DiffLoss(
|
96 |
+
target_channels=self.token_embed_dim,
|
97 |
+
z_channels=decoder_embed_dim,
|
98 |
+
width=diffloss_w,
|
99 |
+
depth=diffloss_d,
|
100 |
+
num_sampling_steps=num_sampling_steps,
|
101 |
+
grad_checkpointing=grad_checkpointing
|
102 |
+
)
|
103 |
+
self.diffusion_batch_mul = diffusion_batch_mul
|
104 |
+
|
105 |
+
def initialize_weights(self):
|
106 |
+
# parameters
|
107 |
+
# torch.nn.init.normal_(self.class_emb.weight, std=.02)
|
108 |
+
# torch.nn.init.normal_(self.fake_latent, std=.02)
|
109 |
+
torch.nn.init.normal_(self.mask_token, std=.02)
|
110 |
+
torch.nn.init.normal_(self.encoder_pos_embed_learned, std=.02)
|
111 |
+
torch.nn.init.normal_(self.decoder_pos_embed_learned, std=.02)
|
112 |
+
torch.nn.init.normal_(self.diffusion_pos_embed_learned, std=.02)
|
113 |
+
|
114 |
+
# initialize nn.Linear and nn.LayerNorm
|
115 |
+
self.apply(self._init_weights)
|
116 |
+
|
117 |
+
def _init_weights(self, m):
|
118 |
+
if isinstance(m, nn.Linear):
|
119 |
+
# we use xavier_uniform following official JAX ViT:
|
120 |
+
torch.nn.init.xavier_uniform_(m.weight)
|
121 |
+
if isinstance(m, nn.Linear) and m.bias is not None:
|
122 |
+
nn.init.constant_(m.bias, 0)
|
123 |
+
elif isinstance(m, nn.LayerNorm):
|
124 |
+
if m.bias is not None:
|
125 |
+
nn.init.constant_(m.bias, 0)
|
126 |
+
if m.weight is not None:
|
127 |
+
nn.init.constant_(m.weight, 1.0)
|
128 |
+
|
129 |
+
def patchify(self, x):
|
130 |
+
bsz, c, h, w = x.shape
|
131 |
+
p = self.patch_size
|
132 |
+
h_, w_ = h // p, w // p
|
133 |
+
|
134 |
+
x = x.reshape(bsz, c, h_, p, w_, p)
|
135 |
+
x = torch.einsum('nchpwq->nhwcpq', x)
|
136 |
+
x = x.reshape(bsz, h_ * w_, c * p ** 2)
|
137 |
+
return x # [n, l, d]
|
138 |
+
|
139 |
+
def unpatchify(self, x):
|
140 |
+
bsz = x.shape[0]
|
141 |
+
p = self.patch_size
|
142 |
+
c = self.vae_embed_dim
|
143 |
+
h_, w_ = self.seq_h, self.seq_w
|
144 |
+
|
145 |
+
x = x.reshape(bsz, h_, w_, c, p, p)
|
146 |
+
x = torch.einsum('nhwcpq->nchpwq', x)
|
147 |
+
x = x.reshape(bsz, c, h_ * p, w_ * p)
|
148 |
+
return x # [n, c, h, w]
|
149 |
+
|
150 |
+
def sample_orders(self, bsz):
|
151 |
+
# generate a batch of random generation orders
|
152 |
+
orders = []
|
153 |
+
for _ in range(bsz):
|
154 |
+
order = np.array(list(range(self.seq_len)))
|
155 |
+
np.random.shuffle(order)
|
156 |
+
orders.append(order)
|
157 |
+
orders = torch.Tensor(np.array(orders)).cuda().long()
|
158 |
+
return orders
|
159 |
+
|
160 |
+
def random_masking(self, x, orders):
|
161 |
+
# generate token mask
|
162 |
+
bsz, seq_len, embed_dim = x.shape
|
163 |
+
mask_rate = self.mask_ratio_generator.rvs(1)[0]
|
164 |
+
num_masked_tokens = int(np.ceil(seq_len * mask_rate))
|
165 |
+
mask = torch.zeros(bsz, seq_len, device=x.device)
|
166 |
+
mask = torch.scatter(mask, dim=-1, index=orders[:, :num_masked_tokens],
|
167 |
+
src=torch.ones(bsz, seq_len, device=x.device))
|
168 |
+
return mask
|
169 |
+
|
170 |
+
def forward_mae_encoder(self, x, mask, y):
|
171 |
+
x = self.z_proj1(x)
|
172 |
+
y = self.z_proj2(y)
|
173 |
+
bsz, seq_len, embed_dim = y.shape
|
174 |
+
|
175 |
+
# concat buffer
|
176 |
+
x = torch.cat([x, y], dim=1)
|
177 |
+
mask_with_buffer = mask #torch.cat([torch.zeros(y.size(0), self.seq_len, device=y.device), mask], dim=1)
|
178 |
+
|
179 |
+
# # random drop class embedding during training
|
180 |
+
# if self.training:
|
181 |
+
# drop_latent_mask = torch.rand(bsz) < self.label_drop_prob
|
182 |
+
# drop_latent_mask = drop_latent_mask.unsqueeze(-1).cuda().to(x.dtype)
|
183 |
+
# class_embedding = drop_latent_mask * self.fake_latent + (1 - drop_latent_mask) * class_embedding
|
184 |
+
|
185 |
+
# x[:, :self.buffer_size] = class_embedding.unsqueeze(1)
|
186 |
+
|
187 |
+
# encoder position embedding
|
188 |
+
x = x + self.encoder_pos_embed_learned
|
189 |
+
x = self.z_proj_ln(x)
|
190 |
+
|
191 |
+
# dropping
|
192 |
+
x = x[(1-mask_with_buffer).nonzero(as_tuple=True)].reshape(bsz, -1, embed_dim)
|
193 |
+
|
194 |
+
# apply Transformer blocks
|
195 |
+
if self.grad_checkpointing and not torch.jit.is_scripting():
|
196 |
+
for block in self.encoder_blocks:
|
197 |
+
x = checkpoint(block, x)
|
198 |
+
else:
|
199 |
+
for block in self.encoder_blocks:
|
200 |
+
x = block(x)
|
201 |
+
x = self.encoder_norm(x)
|
202 |
+
|
203 |
+
return x
|
204 |
+
|
205 |
+
def forward_mae_decoder(self, x, mask):
|
206 |
+
|
207 |
+
x = self.decoder_embed(x)
|
208 |
+
mask_with_buffer = mask#cleartorch.cat([torch.zeros(x.size(0), self.seq_len, device=x.device), mask], dim=1)
|
209 |
+
|
210 |
+
# pad mask tokens
|
211 |
+
mask_tokens = self.mask_token.repeat(mask_with_buffer.shape[0], mask_with_buffer.shape[1], 1).to(x.dtype)
|
212 |
+
x_after_pad = mask_tokens.clone()
|
213 |
+
x_after_pad[(1 - mask_with_buffer).nonzero(as_tuple=True)] = x.reshape(x.shape[0] * x.shape[1], x.shape[2])
|
214 |
+
|
215 |
+
# decoder position embedding
|
216 |
+
x = x_after_pad + self.decoder_pos_embed_learned
|
217 |
+
|
218 |
+
# apply Transformer blocks
|
219 |
+
if self.grad_checkpointing and not torch.jit.is_scripting():
|
220 |
+
for block in self.decoder_blocks:
|
221 |
+
x = checkpoint(block, x)
|
222 |
+
else:
|
223 |
+
for block in self.decoder_blocks:
|
224 |
+
x = block(x)
|
225 |
+
x = self.decoder_norm(x)
|
226 |
+
|
227 |
+
# x = x [:, self.seq_len:]
|
228 |
+
x = x + self.diffusion_pos_embed_learned
|
229 |
+
return x
|
230 |
+
|
231 |
+
def forward_loss(self, z, target, mask):
|
232 |
+
bsz, seq_len, _ = target.shape
|
233 |
+
target = target.reshape(bsz * seq_len, -1).repeat(self.diffusion_batch_mul, 1)
|
234 |
+
z = z.reshape(bsz*seq_len, -1).repeat(self.diffusion_batch_mul, 1)
|
235 |
+
mask = mask.reshape(bsz*seq_len).repeat(self.diffusion_batch_mul)
|
236 |
+
loss = self.diffloss(z=z, target=target, mask=mask)
|
237 |
+
return loss
|
238 |
+
|
239 |
+
def forward(self, imgs, labels):
|
240 |
+
|
241 |
+
# class embed
|
242 |
+
# class_embedding = self.class_emb(labels)
|
243 |
+
|
244 |
+
# patchify and mask (drop) tokens
|
245 |
+
x = self.patchify(imgs)
|
246 |
+
y = self.patchify(labels)
|
247 |
+
gt_latents = torch.cat([x, y], dim=1).clone().detach()
|
248 |
+
orders = self.sample_orders(bsz=y.size(0))
|
249 |
+
mask = self.random_masking(x, orders)
|
250 |
+
mask = torch.cat([torch.zeros(y.size(0), self.seq_len).cuda(), mask], dim=1)
|
251 |
+
# mask = torch.cat([torch.zeros(y.size(0), self.seq_len), torch.ones(y.size(0), self.seq_len)], dim=1)
|
252 |
+
|
253 |
+
# mae encoder
|
254 |
+
x = self.forward_mae_encoder(x, mask, y)
|
255 |
+
|
256 |
+
# mae decoder
|
257 |
+
z = self.forward_mae_decoder(x, mask)
|
258 |
+
|
259 |
+
# diffloss
|
260 |
+
loss = self.forward_loss(z=z, target=gt_latents, mask=mask)
|
261 |
+
|
262 |
+
return loss
|
263 |
+
|
264 |
+
def sample_tokens(self, bsz, num_iter=64, cfg=1.0, cfg_schedule="linear", labels=None, temperature=1.0, progress=False):
|
265 |
+
|
266 |
+
# init and sample generation orders
|
267 |
+
mask = torch.ones(bsz, self.seq_len).cuda()
|
268 |
+
tokens = torch.zeros(bsz, self.seq_len, self.token_embed_dim).cuda()
|
269 |
+
orders = self.sample_orders(bsz)
|
270 |
+
|
271 |
+
indices = list(range(num_iter))
|
272 |
+
if progress:
|
273 |
+
indices = tqdm(indices)
|
274 |
+
# generate latents
|
275 |
+
for step in indices:
|
276 |
+
cur_tokens = tokens.clone()
|
277 |
+
|
278 |
+
# class embedding and CFG
|
279 |
+
if labels is not None:
|
280 |
+
class_embedding = self.class_emb(labels)
|
281 |
+
else:
|
282 |
+
class_embedding = self.fake_latent.repeat(bsz, 1)
|
283 |
+
if not cfg == 1.0:
|
284 |
+
tokens = torch.cat([tokens, tokens], dim=0)
|
285 |
+
class_embedding = torch.cat([class_embedding, self.fake_latent.repeat(bsz, 1)], dim=0)
|
286 |
+
mask = torch.cat([mask, mask], dim=0)
|
287 |
+
|
288 |
+
# mae encoder
|
289 |
+
x = self.forward_mae_encoder(tokens, mask, class_embedding)
|
290 |
+
|
291 |
+
# mae decoder
|
292 |
+
z = self.forward_mae_decoder(x, mask)
|
293 |
+
|
294 |
+
# mask ratio for the next round, following MaskGIT and MAGE.
|
295 |
+
mask_ratio = np.cos(math.pi / 2. * (step + 1) / num_iter)
|
296 |
+
mask_len = torch.Tensor([np.floor(self.seq_len * mask_ratio)]).cuda()
|
297 |
+
|
298 |
+
# masks out at least one for the next iteration
|
299 |
+
mask_len = torch.maximum(torch.Tensor([1]).cuda(),
|
300 |
+
torch.minimum(torch.sum(mask, dim=-1, keepdims=True) - 1, mask_len))
|
301 |
+
|
302 |
+
# get masking for next iteration and locations to be predicted in this iteration
|
303 |
+
mask_next = mask_by_order(mask_len[0], orders, bsz, self.seq_len)
|
304 |
+
if step >= num_iter - 1:
|
305 |
+
mask_to_pred = mask[:bsz].bool()
|
306 |
+
else:
|
307 |
+
mask_to_pred = torch.logical_xor(mask[:bsz].bool(), mask_next.bool())
|
308 |
+
mask = mask_next
|
309 |
+
if not cfg == 1.0:
|
310 |
+
mask_to_pred = torch.cat([mask_to_pred, mask_to_pred], dim=0)
|
311 |
+
|
312 |
+
# sample token latents for this step
|
313 |
+
z = z[mask_to_pred.nonzero(as_tuple=True)]
|
314 |
+
# cfg schedule follow Muse
|
315 |
+
if cfg_schedule == "linear":
|
316 |
+
cfg_iter = 1 + (cfg - 1) * (self.seq_len - mask_len[0]) / self.seq_len
|
317 |
+
elif cfg_schedule == "constant":
|
318 |
+
cfg_iter = cfg
|
319 |
+
else:
|
320 |
+
raise NotImplementedError
|
321 |
+
sampled_token_latent = self.diffloss.sample(z, temperature, cfg_iter)
|
322 |
+
if not cfg == 1.0:
|
323 |
+
sampled_token_latent, _ = sampled_token_latent.chunk(2, dim=0) # Remove null class samples
|
324 |
+
mask_to_pred, _ = mask_to_pred.chunk(2, dim=0)
|
325 |
+
|
326 |
+
cur_tokens[mask_to_pred.nonzero(as_tuple=True)] = sampled_token_latent
|
327 |
+
tokens = cur_tokens.clone()
|
328 |
+
|
329 |
+
# unpatchify
|
330 |
+
tokens = self.unpatchify(tokens)
|
331 |
+
return tokens
|
332 |
+
|
333 |
+
|
334 |
+
def mar_base(**kwargs):
|
335 |
+
model = MAR(
|
336 |
+
encoder_embed_dim=768, encoder_depth=12, encoder_num_heads=12,
|
337 |
+
decoder_embed_dim=768, decoder_depth=12, decoder_num_heads=12,
|
338 |
+
mlp_ratio=4, norm_layer=partial(nn.LayerNorm, eps=1e-6), **kwargs)
|
339 |
+
return model
|
340 |
+
|
341 |
+
|
342 |
+
def mar_large(**kwargs):
|
343 |
+
model = MAR(
|
344 |
+
encoder_embed_dim=1024, encoder_depth=16, encoder_num_heads=16,
|
345 |
+
decoder_embed_dim=1024, decoder_depth=16, decoder_num_heads=16,
|
346 |
+
mlp_ratio=4, norm_layer=partial(nn.LayerNorm, eps=1e-6), **kwargs)
|
347 |
+
return model
|
348 |
+
|
349 |
+
|
350 |
+
def mar_huge(**kwargs):
|
351 |
+
model = MAR(
|
352 |
+
encoder_embed_dim=1280, encoder_depth=20, encoder_num_heads=16,
|
353 |
+
decoder_embed_dim=1280, decoder_depth=20, decoder_num_heads=16,
|
354 |
+
mlp_ratio=4, norm_layer=partial(nn.LayerNorm, eps=1e-6), **kwargs)
|
355 |
+
return model
|
models/vae.py
ADDED
@@ -0,0 +1,67 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import torch
|
2 |
+
import torch.nn as nn
|
3 |
+
import torch.nn.functional as F
|
4 |
+
|
5 |
+
|
6 |
+
from ldm.modules.diffusionmodules.model import Encoder, Decoder
|
7 |
+
from ldm.modules.distributions.distributions import DiagonalGaussianDistribution
|
8 |
+
|
9 |
+
from ldm.util import instantiate_from_config
|
10 |
+
|
11 |
+
|
12 |
+
class AutoencoderKL(nn.Module):
|
13 |
+
def __init__(self,
|
14 |
+
ddconfig,
|
15 |
+
embed_dim,
|
16 |
+
ckpt_path=None,
|
17 |
+
ignore_keys=[],
|
18 |
+
image_key="image",
|
19 |
+
colorize_nlabels=None,
|
20 |
+
monitor=None,
|
21 |
+
):
|
22 |
+
super().__init__()
|
23 |
+
self.image_key = image_key
|
24 |
+
self.encoder = Encoder(**ddconfig)
|
25 |
+
self.decoder = Decoder(**ddconfig)
|
26 |
+
assert ddconfig["double_z"]
|
27 |
+
self.quant_conv = torch.nn.Conv2d(2*ddconfig["z_channels"], 2*embed_dim, 1)
|
28 |
+
self.post_quant_conv = torch.nn.Conv2d(embed_dim, ddconfig["z_channels"], 1)
|
29 |
+
self.embed_dim = embed_dim
|
30 |
+
if colorize_nlabels is not None:
|
31 |
+
assert type(colorize_nlabels)==int
|
32 |
+
self.register_buffer("colorize", torch.randn(3, colorize_nlabels, 1, 1))
|
33 |
+
if monitor is not None:
|
34 |
+
self.monitor = monitor
|
35 |
+
if ckpt_path is not None:
|
36 |
+
self.init_from_ckpt(ckpt_path, ignore_keys=ignore_keys)
|
37 |
+
|
38 |
+
def init_from_ckpt(self, path, ignore_keys=list()):
|
39 |
+
sd = torch.load(path, map_location="cpu")["state_dict"]
|
40 |
+
keys = list(sd.keys())
|
41 |
+
for k in keys:
|
42 |
+
for ik in ignore_keys:
|
43 |
+
if k.startswith(ik):
|
44 |
+
print("Deleting key {} from state_dict.".format(k))
|
45 |
+
del sd[k]
|
46 |
+
self.load_state_dict(sd, strict=False)
|
47 |
+
print(f"Restored from {path}")
|
48 |
+
|
49 |
+
def encode(self, x):
|
50 |
+
h = self.encoder(x)
|
51 |
+
moments = self.quant_conv(h)
|
52 |
+
posterior = DiagonalGaussianDistribution(moments)
|
53 |
+
return posterior
|
54 |
+
|
55 |
+
def decode(self, z):
|
56 |
+
z = self.post_quant_conv(z)
|
57 |
+
dec = self.decoder(z)
|
58 |
+
return dec
|
59 |
+
|
60 |
+
def forward(self, input, sample_posterior=True):
|
61 |
+
posterior = self.encode(input)
|
62 |
+
if sample_posterior:
|
63 |
+
z = posterior.sample()
|
64 |
+
else:
|
65 |
+
z = posterior.mode()
|
66 |
+
dec = self.decode(z)
|
67 |
+
return dec, posterior
|
taming/modules/autoencoder/lpips/vgg.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:a78928a0af1e5f0fcb1f3b9e8f8c3a2a5a3de244d830ad5c1feddc79b8432868
|
3 |
+
size 7289
|
util/crop.py
ADDED
@@ -0,0 +1,23 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import numpy as np
|
2 |
+
from PIL import Image
|
3 |
+
|
4 |
+
|
5 |
+
def center_crop_arr(pil_image, image_size):
|
6 |
+
"""
|
7 |
+
Center cropping implementation from ADM.
|
8 |
+
https://github.com/openai/guided-diffusion/blob/8fb3ad9197f16bbc40620447b2742e13458d2831/guided_diffusion/image_datasets.py#L126
|
9 |
+
"""
|
10 |
+
while min(*pil_image.size) >= 2 * image_size:
|
11 |
+
pil_image = pil_image.resize(
|
12 |
+
tuple(x // 2 for x in pil_image.size), resample=Image.BOX
|
13 |
+
)
|
14 |
+
|
15 |
+
scale = image_size / min(*pil_image.size)
|
16 |
+
pil_image = pil_image.resize(
|
17 |
+
tuple(round(x * scale) for x in pil_image.size), resample=Image.BICUBIC
|
18 |
+
)
|
19 |
+
|
20 |
+
arr = np.array(pil_image)
|
21 |
+
crop_y = (arr.shape[0] - image_size) // 2
|
22 |
+
crop_x = (arr.shape[1] - image_size) // 2
|
23 |
+
return Image.fromarray(arr[crop_y: crop_y + image_size, crop_x: crop_x + image_size])
|
util/download.py
ADDED
@@ -0,0 +1,62 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
from tqdm import tqdm
|
3 |
+
import requests
|
4 |
+
|
5 |
+
|
6 |
+
def download_pretrained_vae(overwrite=False):
|
7 |
+
download_path = "pretrained_models/vae/kl16.ckpt"
|
8 |
+
if not os.path.exists(download_path) or overwrite:
|
9 |
+
headers = {'user-agent': 'Wget/1.16 (linux-gnu)'}
|
10 |
+
os.makedirs("pretrained_models/vae", exist_ok=True)
|
11 |
+
r = requests.get("https://www.dropbox.com/scl/fi/hhmuvaiacrarfg28qxhwz/kl16.ckpt?rlkey=l44xipsezc8atcffdp4q7mwmh&dl=0", stream=True, headers=headers)
|
12 |
+
print("Downloading KL-16 VAE...")
|
13 |
+
with open(download_path, 'wb') as f:
|
14 |
+
for chunk in tqdm(r.iter_content(chunk_size=1024*1024), unit="MB", total=254):
|
15 |
+
if chunk:
|
16 |
+
f.write(chunk)
|
17 |
+
|
18 |
+
|
19 |
+
def download_pretrained_marb(overwrite=False):
|
20 |
+
download_path = "pretrained_models/mar/mar_base/checkpoint-last.pth"
|
21 |
+
if not os.path.exists(download_path) or overwrite:
|
22 |
+
headers = {'user-agent': 'Wget/1.16 (linux-gnu)'}
|
23 |
+
os.makedirs("pretrained_models/mar/mar_base", exist_ok=True)
|
24 |
+
r = requests.get("https://www.dropbox.com/scl/fi/f6dpuyjb7fudzxcyhvrhk/checkpoint-last.pth?rlkey=a6i4bo71vhfo4anp33n9ukujb&dl=0", stream=True, headers=headers)
|
25 |
+
print("Downloading MAR-B...")
|
26 |
+
with open(download_path, 'wb') as f:
|
27 |
+
for chunk in tqdm(r.iter_content(chunk_size=1024*1024), unit="MB", total=1587):
|
28 |
+
if chunk:
|
29 |
+
f.write(chunk)
|
30 |
+
|
31 |
+
|
32 |
+
def download_pretrained_marl(overwrite=False):
|
33 |
+
download_path = "pretrained_models/mar/mar_large/checkpoint-last.pth"
|
34 |
+
if not os.path.exists(download_path) or overwrite:
|
35 |
+
headers = {'user-agent': 'Wget/1.16 (linux-gnu)'}
|
36 |
+
os.makedirs("pretrained_models/mar/mar_large", exist_ok=True)
|
37 |
+
r = requests.get("https://www.dropbox.com/scl/fi/pxacc5b2mrt3ifw4cah6k/checkpoint-last.pth?rlkey=m48ovo6g7ivcbosrbdaz0ehqt&dl=0", stream=True, headers=headers)
|
38 |
+
print("Downloading MAR-L...")
|
39 |
+
with open(download_path, 'wb') as f:
|
40 |
+
for chunk in tqdm(r.iter_content(chunk_size=1024*1024), unit="MB", total=3650):
|
41 |
+
if chunk:
|
42 |
+
f.write(chunk)
|
43 |
+
|
44 |
+
|
45 |
+
def download_pretrained_marh(overwrite=False):
|
46 |
+
download_path = "pretrained_models/mar/mar_huge/checkpoint-last.pth"
|
47 |
+
if not os.path.exists(download_path) or overwrite:
|
48 |
+
headers = {'user-agent': 'Wget/1.16 (linux-gnu)'}
|
49 |
+
os.makedirs("pretrained_models/mar/mar_huge", exist_ok=True)
|
50 |
+
r = requests.get("https://www.dropbox.com/scl/fi/1qmfx6fpy3k7j9vcjjs3s/checkpoint-last.pth?rlkey=4lae281yzxb406atp32vzc83o&dl=0", stream=True, headers=headers)
|
51 |
+
print("Downloading MAR-H...")
|
52 |
+
with open(download_path, 'wb') as f:
|
53 |
+
for chunk in tqdm(r.iter_content(chunk_size=1024*1024), unit="MB", total=7191):
|
54 |
+
if chunk:
|
55 |
+
f.write(chunk)
|
56 |
+
|
57 |
+
|
58 |
+
if __name__ == "__main__":
|
59 |
+
download_pretrained_vae()
|
60 |
+
download_pretrained_marb()
|
61 |
+
download_pretrained_marl()
|
62 |
+
download_pretrained_marh()
|
util/loader.py
ADDED
@@ -0,0 +1,56 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
import numpy as np
|
3 |
+
|
4 |
+
import torch
|
5 |
+
import torchvision.datasets as datasets
|
6 |
+
|
7 |
+
|
8 |
+
class ImageFolderWithFilename(datasets.ImageFolder):
|
9 |
+
def __getitem__(self, index: int):
|
10 |
+
"""
|
11 |
+
Args:
|
12 |
+
index (int): Index
|
13 |
+
|
14 |
+
Returns:
|
15 |
+
tuple: (sample, target, filename).
|
16 |
+
"""
|
17 |
+
path, target = self.samples[index]
|
18 |
+
sample = self.loader(path)
|
19 |
+
if self.transform is not None:
|
20 |
+
sample = self.transform(sample)
|
21 |
+
if self.target_transform is not None:
|
22 |
+
target = self.target_transform(target)
|
23 |
+
|
24 |
+
filename = path.split(os.path.sep)[-2:]
|
25 |
+
filename = os.path.join(*filename)
|
26 |
+
return sample, target, filename
|
27 |
+
|
28 |
+
|
29 |
+
class CachedFolder(datasets.DatasetFolder):
|
30 |
+
def __init__(
|
31 |
+
self,
|
32 |
+
root: str,
|
33 |
+
):
|
34 |
+
super().__init__(
|
35 |
+
root,
|
36 |
+
loader=None,
|
37 |
+
extensions=(".npz",),
|
38 |
+
)
|
39 |
+
|
40 |
+
def __getitem__(self, index: int):
|
41 |
+
"""
|
42 |
+
Args:
|
43 |
+
index (int): Index
|
44 |
+
|
45 |
+
Returns:
|
46 |
+
tuple: (moments, target).
|
47 |
+
"""
|
48 |
+
path, target = self.samples[index]
|
49 |
+
|
50 |
+
data = np.load(path)
|
51 |
+
if torch.rand(1) < 0.5: # randomly hflip
|
52 |
+
moments = data['moments']
|
53 |
+
else:
|
54 |
+
moments = data['moments_flip']
|
55 |
+
|
56 |
+
return moments, target
|
util/lr_sched.py
ADDED
@@ -0,0 +1,21 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import math
|
2 |
+
|
3 |
+
|
4 |
+
def adjust_learning_rate(optimizer, epoch, args):
|
5 |
+
"""Decay the learning rate with half-cycle cosine after warmup"""
|
6 |
+
if epoch < args.warmup_epochs:
|
7 |
+
lr = args.lr * epoch / args.warmup_epochs
|
8 |
+
else:
|
9 |
+
if args.lr_schedule == "constant":
|
10 |
+
lr = args.lr
|
11 |
+
elif args.lr_schedule == "cosine":
|
12 |
+
lr = args.min_lr + (args.lr - args.min_lr) * 0.5 * \
|
13 |
+
(1. + math.cos(math.pi * (epoch - args.warmup_epochs) / (args.epochs - args.warmup_epochs)))
|
14 |
+
else:
|
15 |
+
raise NotImplementedError
|
16 |
+
for param_group in optimizer.param_groups:
|
17 |
+
if "lr_scale" in param_group:
|
18 |
+
param_group["lr"] = lr * param_group["lr_scale"]
|
19 |
+
else:
|
20 |
+
param_group["lr"] = lr
|
21 |
+
return lr
|
util/misc.py
ADDED
@@ -0,0 +1,340 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import builtins
|
2 |
+
import datetime
|
3 |
+
import os
|
4 |
+
import time
|
5 |
+
from collections import defaultdict, deque
|
6 |
+
from pathlib import Path
|
7 |
+
|
8 |
+
import torch
|
9 |
+
import torch.distributed as dist
|
10 |
+
TORCH_MAJOR = int(torch.__version__.split('.')[0])
|
11 |
+
TORCH_MINOR = int(torch.__version__.split('.')[1])
|
12 |
+
|
13 |
+
if TORCH_MAJOR == 1 and TORCH_MINOR < 8:
|
14 |
+
from torch._six import inf
|
15 |
+
else:
|
16 |
+
from torch import inf
|
17 |
+
import copy
|
18 |
+
|
19 |
+
|
20 |
+
class SmoothedValue(object):
|
21 |
+
"""Track a series of values and provide access to smoothed values over a
|
22 |
+
window or the global series average.
|
23 |
+
"""
|
24 |
+
|
25 |
+
def __init__(self, window_size=20, fmt=None):
|
26 |
+
if fmt is None:
|
27 |
+
fmt = "{median:.4f} ({global_avg:.4f})"
|
28 |
+
self.deque = deque(maxlen=window_size)
|
29 |
+
self.total = 0.0
|
30 |
+
self.count = 0
|
31 |
+
self.fmt = fmt
|
32 |
+
|
33 |
+
def update(self, value, n=1):
|
34 |
+
self.deque.append(value)
|
35 |
+
self.count += n
|
36 |
+
self.total += value * n
|
37 |
+
|
38 |
+
def synchronize_between_processes(self):
|
39 |
+
"""
|
40 |
+
Warning: does not synchronize the deque!
|
41 |
+
"""
|
42 |
+
if not is_dist_avail_and_initialized():
|
43 |
+
return
|
44 |
+
t = torch.tensor([self.count, self.total], dtype=torch.float64, device='cuda')
|
45 |
+
dist.barrier()
|
46 |
+
dist.all_reduce(t)
|
47 |
+
t = t.tolist()
|
48 |
+
self.count = int(t[0])
|
49 |
+
self.total = t[1]
|
50 |
+
|
51 |
+
@property
|
52 |
+
def median(self):
|
53 |
+
d = torch.tensor(list(self.deque))
|
54 |
+
return d.median().item()
|
55 |
+
|
56 |
+
@property
|
57 |
+
def avg(self):
|
58 |
+
d = torch.tensor(list(self.deque), dtype=torch.float32)
|
59 |
+
return d.mean().item()
|
60 |
+
|
61 |
+
@property
|
62 |
+
def global_avg(self):
|
63 |
+
return self.total / self.count
|
64 |
+
|
65 |
+
@property
|
66 |
+
def max(self):
|
67 |
+
return max(self.deque)
|
68 |
+
|
69 |
+
@property
|
70 |
+
def value(self):
|
71 |
+
return self.deque[-1]
|
72 |
+
|
73 |
+
def __str__(self):
|
74 |
+
return self.fmt.format(
|
75 |
+
median=self.median,
|
76 |
+
avg=self.avg,
|
77 |
+
global_avg=self.global_avg,
|
78 |
+
max=self.max,
|
79 |
+
value=self.value)
|
80 |
+
|
81 |
+
|
82 |
+
class MetricLogger(object):
|
83 |
+
def __init__(self, delimiter="\t"):
|
84 |
+
self.meters = defaultdict(SmoothedValue)
|
85 |
+
self.delimiter = delimiter
|
86 |
+
|
87 |
+
def update(self, **kwargs):
|
88 |
+
for k, v in kwargs.items():
|
89 |
+
if v is None:
|
90 |
+
continue
|
91 |
+
if isinstance(v, torch.Tensor):
|
92 |
+
v = v.item()
|
93 |
+
assert isinstance(v, (float, int))
|
94 |
+
self.meters[k].update(v)
|
95 |
+
|
96 |
+
def __getattr__(self, attr):
|
97 |
+
if attr in self.meters:
|
98 |
+
return self.meters[attr]
|
99 |
+
if attr in self.__dict__:
|
100 |
+
return self.__dict__[attr]
|
101 |
+
raise AttributeError("'{}' object has no attribute '{}'".format(
|
102 |
+
type(self).__name__, attr))
|
103 |
+
|
104 |
+
def __str__(self):
|
105 |
+
loss_str = []
|
106 |
+
for name, meter in self.meters.items():
|
107 |
+
loss_str.append(
|
108 |
+
"{}: {}".format(name, str(meter))
|
109 |
+
)
|
110 |
+
return self.delimiter.join(loss_str)
|
111 |
+
|
112 |
+
def synchronize_between_processes(self):
|
113 |
+
for meter in self.meters.values():
|
114 |
+
meter.synchronize_between_processes()
|
115 |
+
|
116 |
+
def add_meter(self, name, meter):
|
117 |
+
self.meters[name] = meter
|
118 |
+
|
119 |
+
def log_every(self, iterable, print_freq, header=None):
|
120 |
+
i = 0
|
121 |
+
if not header:
|
122 |
+
header = ''
|
123 |
+
start_time = time.time()
|
124 |
+
end = time.time()
|
125 |
+
iter_time = SmoothedValue(fmt='{avg:.4f}')
|
126 |
+
data_time = SmoothedValue(fmt='{avg:.4f}')
|
127 |
+
space_fmt = ':' + str(len(str(len(iterable)))) + 'd'
|
128 |
+
log_msg = [
|
129 |
+
header,
|
130 |
+
'[{0' + space_fmt + '}/{1}]',
|
131 |
+
'eta: {eta}',
|
132 |
+
'{meters}',
|
133 |
+
'time: {time}',
|
134 |
+
'data: {data}'
|
135 |
+
]
|
136 |
+
if torch.cuda.is_available():
|
137 |
+
log_msg.append('max mem: {memory:.0f}')
|
138 |
+
log_msg = self.delimiter.join(log_msg)
|
139 |
+
MB = 1024.0 * 1024.0
|
140 |
+
for obj in iterable:
|
141 |
+
data_time.update(time.time() - end)
|
142 |
+
yield obj
|
143 |
+
iter_time.update(time.time() - end)
|
144 |
+
if i % print_freq == 0 or i == len(iterable) - 1:
|
145 |
+
eta_seconds = iter_time.global_avg * (len(iterable) - i)
|
146 |
+
eta_string = str(datetime.timedelta(seconds=int(eta_seconds)))
|
147 |
+
if torch.cuda.is_available():
|
148 |
+
print(log_msg.format(
|
149 |
+
i, len(iterable), eta=eta_string,
|
150 |
+
meters=str(self),
|
151 |
+
time=str(iter_time), data=str(data_time),
|
152 |
+
memory=torch.cuda.max_memory_allocated() / MB))
|
153 |
+
else:
|
154 |
+
print(log_msg.format(
|
155 |
+
i, len(iterable), eta=eta_string,
|
156 |
+
meters=str(self),
|
157 |
+
time=str(iter_time), data=str(data_time)))
|
158 |
+
i += 1
|
159 |
+
end = time.time()
|
160 |
+
total_time = time.time() - start_time
|
161 |
+
total_time_str = str(datetime.timedelta(seconds=int(total_time)))
|
162 |
+
print('{} Total time: {} ({:.4f} s / it)'.format(
|
163 |
+
header, total_time_str, total_time / len(iterable)))
|
164 |
+
|
165 |
+
|
166 |
+
def setup_for_distributed(is_master):
|
167 |
+
"""
|
168 |
+
This function disables printing when not in master process
|
169 |
+
"""
|
170 |
+
builtin_print = builtins.print
|
171 |
+
|
172 |
+
def print(*args, **kwargs):
|
173 |
+
force = kwargs.pop('force', False)
|
174 |
+
force = force or (get_world_size() > 8)
|
175 |
+
if is_master or force:
|
176 |
+
now = datetime.datetime.now().time()
|
177 |
+
builtin_print('[{}] '.format(now), end='') # print with time stamp
|
178 |
+
builtin_print(*args, **kwargs)
|
179 |
+
|
180 |
+
builtins.print = print
|
181 |
+
|
182 |
+
|
183 |
+
def is_dist_avail_and_initialized():
|
184 |
+
if not dist.is_available():
|
185 |
+
return False
|
186 |
+
if not dist.is_initialized():
|
187 |
+
return False
|
188 |
+
return True
|
189 |
+
|
190 |
+
|
191 |
+
def get_world_size():
|
192 |
+
if not is_dist_avail_and_initialized():
|
193 |
+
return 1
|
194 |
+
return dist.get_world_size()
|
195 |
+
|
196 |
+
|
197 |
+
def get_rank():
|
198 |
+
if not is_dist_avail_and_initialized():
|
199 |
+
return 0
|
200 |
+
return dist.get_rank()
|
201 |
+
|
202 |
+
|
203 |
+
def is_main_process():
|
204 |
+
return get_rank() == 0
|
205 |
+
|
206 |
+
|
207 |
+
def save_on_master(*args, **kwargs):
|
208 |
+
if is_main_process():
|
209 |
+
torch.save(*args, **kwargs)
|
210 |
+
|
211 |
+
|
212 |
+
def init_distributed_mode(args):
|
213 |
+
if args.dist_on_itp:
|
214 |
+
args.rank = int(os.environ['OMPI_COMM_WORLD_RANK'])
|
215 |
+
args.world_size = int(os.environ['OMPI_COMM_WORLD_SIZE'])
|
216 |
+
args.gpu = int(os.environ['OMPI_COMM_WORLD_LOCAL_RANK'])
|
217 |
+
args.dist_url = "tcp://%s:%s" % (os.environ['MASTER_ADDR'], os.environ['MASTER_PORT'])
|
218 |
+
os.environ['LOCAL_RANK'] = str(args.gpu)
|
219 |
+
os.environ['RANK'] = str(args.rank)
|
220 |
+
os.environ['WORLD_SIZE'] = str(args.world_size)
|
221 |
+
# ["RANK", "WORLD_SIZE", "MASTER_ADDR", "MASTER_PORT", "LOCAL_RANK"]
|
222 |
+
elif 'RANK' in os.environ and 'WORLD_SIZE' in os.environ:
|
223 |
+
args.rank = int(os.environ["RANK"])
|
224 |
+
args.world_size = int(os.environ['WORLD_SIZE'])
|
225 |
+
args.gpu = int(os.environ['LOCAL_RANK'])
|
226 |
+
elif 'SLURM_PROCID' in os.environ:
|
227 |
+
args.rank = int(os.environ['SLURM_PROCID'])
|
228 |
+
args.gpu = args.rank % torch.cuda.device_count()
|
229 |
+
else:
|
230 |
+
print('Not using distributed mode')
|
231 |
+
setup_for_distributed(is_master=True) # hack
|
232 |
+
args.distributed = False
|
233 |
+
return
|
234 |
+
|
235 |
+
args.distributed = True
|
236 |
+
|
237 |
+
torch.cuda.set_device(args.gpu)
|
238 |
+
args.dist_backend = 'nccl'
|
239 |
+
print('| distributed init (rank {}): {}, gpu {}'.format(
|
240 |
+
args.rank, args.dist_url, args.gpu), flush=True)
|
241 |
+
torch.distributed.init_process_group(backend=args.dist_backend, init_method=args.dist_url,
|
242 |
+
world_size=args.world_size, rank=args.rank)
|
243 |
+
torch.distributed.barrier()
|
244 |
+
setup_for_distributed(args.rank == 0)
|
245 |
+
|
246 |
+
|
247 |
+
class NativeScalerWithGradNormCount:
|
248 |
+
state_dict_key = "amp_scaler"
|
249 |
+
|
250 |
+
def __init__(self):
|
251 |
+
self._scaler = torch.cuda.amp.GradScaler()
|
252 |
+
|
253 |
+
def __call__(self, loss, optimizer, clip_grad=None, parameters=None, create_graph=False, update_grad=True):
|
254 |
+
self._scaler.scale(loss).backward(create_graph=create_graph)
|
255 |
+
if update_grad:
|
256 |
+
if clip_grad is not None:
|
257 |
+
assert parameters is not None
|
258 |
+
self._scaler.unscale_(optimizer) # unscale the gradients of optimizer's assigned params in-place
|
259 |
+
norm = torch.nn.utils.clip_grad_norm_(parameters, clip_grad)
|
260 |
+
else:
|
261 |
+
self._scaler.unscale_(optimizer)
|
262 |
+
norm = get_grad_norm_(parameters)
|
263 |
+
self._scaler.step(optimizer)
|
264 |
+
self._scaler.update()
|
265 |
+
else:
|
266 |
+
norm = None
|
267 |
+
return norm
|
268 |
+
|
269 |
+
def state_dict(self):
|
270 |
+
return self._scaler.state_dict()
|
271 |
+
|
272 |
+
def load_state_dict(self, state_dict):
|
273 |
+
self._scaler.load_state_dict(state_dict)
|
274 |
+
|
275 |
+
|
276 |
+
def get_grad_norm_(parameters, norm_type: float = 2.0) -> torch.Tensor:
|
277 |
+
if isinstance(parameters, torch.Tensor):
|
278 |
+
parameters = [parameters]
|
279 |
+
parameters = [p for p in parameters if p.grad is not None]
|
280 |
+
norm_type = float(norm_type)
|
281 |
+
if len(parameters) == 0:
|
282 |
+
return torch.tensor(0.)
|
283 |
+
device = parameters[0].grad.device
|
284 |
+
if norm_type == inf:
|
285 |
+
total_norm = max(p.grad.detach().abs().max().to(device) for p in parameters)
|
286 |
+
else:
|
287 |
+
total_norm = torch.norm(torch.stack([torch.norm(p.grad.detach(), norm_type).to(device) for p in parameters]), norm_type)
|
288 |
+
return total_norm
|
289 |
+
|
290 |
+
|
291 |
+
def add_weight_decay(model, weight_decay=1e-5, skip_list=()):
|
292 |
+
decay = []
|
293 |
+
no_decay = []
|
294 |
+
for name, param in model.named_parameters():
|
295 |
+
if not param.requires_grad:
|
296 |
+
continue # frozen weights
|
297 |
+
if len(param.shape) == 1 or name.endswith(".bias") or name in skip_list or 'diffloss' in name:
|
298 |
+
no_decay.append(param) # no weight decay on bias, norm and diffloss
|
299 |
+
else:
|
300 |
+
decay.append(param)
|
301 |
+
return [
|
302 |
+
{'params': no_decay, 'weight_decay': 0.},
|
303 |
+
{'params': decay, 'weight_decay': weight_decay}]
|
304 |
+
|
305 |
+
|
306 |
+
def save_model(args, epoch, model, model_without_ddp, optimizer, loss_scaler, ema_params=None, epoch_name=None):
|
307 |
+
if epoch_name is None:
|
308 |
+
epoch_name = str(epoch)
|
309 |
+
output_dir = Path(args.output_dir)
|
310 |
+
checkpoint_path = output_dir / ('checkpoint-%s.pth' % epoch_name)
|
311 |
+
|
312 |
+
# ema
|
313 |
+
if ema_params is not None:
|
314 |
+
ema_state_dict = copy.deepcopy(model_without_ddp.state_dict())
|
315 |
+
for i, (name, _value) in enumerate(model_without_ddp.named_parameters()):
|
316 |
+
assert name in ema_state_dict
|
317 |
+
ema_state_dict[name] = ema_params[i]
|
318 |
+
else:
|
319 |
+
ema_state_dict = None
|
320 |
+
|
321 |
+
to_save = {
|
322 |
+
'model': model_without_ddp.state_dict(),
|
323 |
+
'model_ema': ema_state_dict,
|
324 |
+
'optimizer': optimizer.state_dict(),
|
325 |
+
'epoch': epoch,
|
326 |
+
'scaler': loss_scaler.state_dict(),
|
327 |
+
'args': args,
|
328 |
+
}
|
329 |
+
save_on_master(to_save, checkpoint_path)
|
330 |
+
|
331 |
+
|
332 |
+
def all_reduce_mean(x):
|
333 |
+
world_size = get_world_size()
|
334 |
+
if world_size > 1:
|
335 |
+
x_reduce = torch.tensor(x).cuda()
|
336 |
+
dist.all_reduce(x_reduce)
|
337 |
+
x_reduce /= world_size
|
338 |
+
return x_reduce.item()
|
339 |
+
else:
|
340 |
+
return x
|