forked from mindspore-lab/mindone
-
Notifications
You must be signed in to change notification settings - Fork 0
/
svd.yaml
157 lines (144 loc) · 4.79 KB
/
svd.yaml
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
model:
target: models.diffusion.VideoDiffusionEngine
params:
scale_factor: 0.18215
disable_first_stage_amp: True
denoiser_config:
target: gm.modules.diffusionmodules.denoiser.Denoiser
params:
weighting_config:
target: gm.modules.diffusionmodules.denoiser_weighting.EDMWeighting
params:
sigma_data: 1.0
scaling_config:
target: gm.modules.diffusionmodules.denoiser_scaling.VScalingWithEDMcNoise
network_config:
target: modules.unet3d.VideoUNet
params:
adm_in_channels: 768
num_classes: sequential
in_channels: 8
out_channels: 4
model_channels: 320
attention_resolutions: [ 4, 2, 1 ]
num_res_blocks: 2
channel_mult: [ 1, 2, 4, 4 ]
num_head_channels: 64
use_linear_in_transformer: True
transformer_depth: 1
context_dim: 1024
spatial_transformer_attn_type: flash-attention
extra_ff_mix_layer: True
use_spatial_context: True
merge_strategy: learned
video_kernel_size: [ 3, 1, 1 ]
conditioner_config:
target: gm.modules.GeneralConditioner
params:
emb_models:
- is_trainable: False
input_key: cond_frames_without_noise
target: modules.encoders.modules.FrozenOpenCLIPImagePredictionEmbedder
params:
n_cond_frames: 1
n_copies: 1
open_clip_embedding_config:
target: gm.modules.embedders.modules.FrozenOpenCLIPImageEmbedder
params:
freeze: True
- input_key: fps_id
is_trainable: False
target: gm.modules.embedders.modules.ConcatTimestepEmbedderND
params:
outdim: 256
- input_key: motion_bucket_id
is_trainable: False
target: gm.modules.embedders.modules.ConcatTimestepEmbedderND
params:
outdim: 256
- input_key: cond_frames
is_trainable: False
target: modules.encoders.modules.VideoPredictionEmbedderWithEncoder
params:
disable_encoder_amp: True
n_cond_frames: 1
n_copies: 1
is_ae: True
encoder_config:
target: gm.models.autoencoder.AutoencoderKLModeOnly
params:
embed_dim: 4
monitor: val/rec_loss
ddconfig:
attn_type: vanilla
double_z: True
z_channels: 4
resolution: 256
in_channels: 3
out_ch: 3
ch: 128
ch_mult: [ 1, 2, 4, 4 ]
num_res_blocks: 2
attn_resolutions: [ ]
dropout: 0.0
- input_key: cond_aug
is_trainable: False
target: gm.modules.embedders.modules.ConcatTimestepEmbedderND
params:
outdim: 256
first_stage_config:
target: gm.models.autoencoder.AutoencodingEngine
params:
loss_config:
target: mindspore.nn.Identity
regularizer_config:
target: gm.modules.autoencoding.regularizers.DiagonalGaussianRegularizer
encoder_config:
target: gm.modules.diffusionmodules.model.Encoder
params:
attn_type: vanilla
double_z: True
z_channels: 4
resolution: 256
in_channels: 3
out_ch: 3
ch: 128
ch_mult: [ 1, 2, 4, 4 ]
num_res_blocks: 2
attn_resolutions: [ ]
dropout: 0.0
decoder_config:
target: modules.temporal_ae.VideoDecoder
params:
attn_type: vanilla
double_z: True
z_channels: 4
resolution: 256
in_channels: 3
out_ch: 3
ch: 128
ch_mult: [ 1, 2, 4, 4 ]
num_res_blocks: 2
attn_resolutions: [ ]
dropout: 0.0
video_kernel_size: [ 3, 1, 1 ]
sigma_sampler_config:
target: gm.modules.diffusionmodules.sigma_sampling.EDMSampling
params:
p_mean: 1.0
p_std: 1.6
loss_fn_config:
target: gm.modules.diffusionmodules.loss.StandardDiffusionLoss
sampler_config:
target: modules.diffusionmodules.sampler.EulerEDMSampler
params:
num_steps: 25
discretization_config:
target: gm.modules.diffusionmodules.discretizer.EDMDiscretization
params:
sigma_max: 700.0
guider_config:
target: modules.diffusionmodules.guiders.LinearPredictionGuider
params:
max_scale: 2.5
min_scale: 1.0