CUMTHenry commited on
Commit
7010e84
·
verified ·
1 Parent(s): 18f8c1c

Upload humanoidmaze-medium-stitch-v0/diffusion/og_humM_Me_o2d_Cd_Stgl_PadBuf_Ft64_ts1k_h336_ovlp128_ovdm5_ts512_bs192_bd128_td96_drop02_T1000/model_config.txt with huggingface_hub

Browse files
humanoidmaze-medium-stitch-v0/diffusion/og_humM_Me_o2d_Cd_Stgl_PadBuf_Ft64_ts1k_h336_ovlp128_ovdm5_ts512_bs192_bd128_td96_drop02_T1000/model_config.txt ADDED
@@ -0,0 +1,1118 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ Unet1D_TjTi_Stgl_Cond_V1(
2
+ (st_ovlp_model): Traj_Time_Encoder(
3
+ (time_mlp): Sequential(
4
+ (0): SinusoidalPosEmb()
5
+ (1): Linear(in_features=32, out_features=128, bias=True)
6
+ (2): Mish()
7
+ (3): Linear(in_features=128, out_features=32, bias=True)
8
+ )
9
+ (downs): ModuleList(
10
+ (0): ModuleList(
11
+ (0): Hi_ResidualTemporalBlock(
12
+ (blocks): ModuleList(
13
+ (0): Conv1dBlock(
14
+ (block): Sequential(
15
+ (0): Conv1d(2, 32, kernel_size=(5,), stride=(1,), padding=(2,))
16
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
17
+ (2): GroupNorm(8, 32, eps=1e-05, affine=True)
18
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
19
+ (4): Mish()
20
+ )
21
+ )
22
+ (1): Conv1dBlock(
23
+ (block): Sequential(
24
+ (0): Conv1d(32, 32, kernel_size=(5,), stride=(1,), padding=(2,))
25
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
26
+ (2): GroupNorm(8, 32, eps=1e-05, affine=True)
27
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
28
+ (4): Mish()
29
+ )
30
+ )
31
+ )
32
+ (time_mlp): Sequential(
33
+ (0): Linear(in_features=32, out_features=64, bias=True)
34
+ (1): Mish()
35
+ (2): Linear(in_features=64, out_features=32, bias=True)
36
+ (3): Rearrange('batch t -> batch t 1')
37
+ )
38
+ (residual_conv): Conv1d(2, 32, kernel_size=(1,), stride=(1,))
39
+ )
40
+ (1): Hi_ResidualTemporalBlock(
41
+ (blocks): ModuleList(
42
+ (0-1): 2 x Conv1dBlock(
43
+ (block): Sequential(
44
+ (0): Conv1d(32, 32, kernel_size=(5,), stride=(1,), padding=(2,))
45
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
46
+ (2): GroupNorm(8, 32, eps=1e-05, affine=True)
47
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
48
+ (4): Mish()
49
+ )
50
+ )
51
+ )
52
+ (time_mlp): Sequential(
53
+ (0): Linear(in_features=32, out_features=64, bias=True)
54
+ (1): Mish()
55
+ (2): Linear(in_features=64, out_features=32, bias=True)
56
+ (3): Rearrange('batch t -> batch t 1')
57
+ )
58
+ (residual_conv): Identity()
59
+ )
60
+ (2): Downsample1d(
61
+ (conv): Conv1d(32, 32, kernel_size=(3,), stride=(2,), padding=(1,))
62
+ )
63
+ )
64
+ (1): ModuleList(
65
+ (0): Hi_ResidualTemporalBlock(
66
+ (blocks): ModuleList(
67
+ (0): Conv1dBlock(
68
+ (block): Sequential(
69
+ (0): Conv1d(32, 64, kernel_size=(5,), stride=(1,), padding=(2,))
70
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
71
+ (2): GroupNorm(8, 64, eps=1e-05, affine=True)
72
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
73
+ (4): Mish()
74
+ )
75
+ )
76
+ (1): Conv1dBlock(
77
+ (block): Sequential(
78
+ (0): Conv1d(64, 64, kernel_size=(5,), stride=(1,), padding=(2,))
79
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
80
+ (2): GroupNorm(8, 64, eps=1e-05, affine=True)
81
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
82
+ (4): Mish()
83
+ )
84
+ )
85
+ )
86
+ (time_mlp): Sequential(
87
+ (0): Linear(in_features=32, out_features=64, bias=True)
88
+ (1): Mish()
89
+ (2): Linear(in_features=64, out_features=64, bias=True)
90
+ (3): Rearrange('batch t -> batch t 1')
91
+ )
92
+ (residual_conv): Conv1d(32, 64, kernel_size=(1,), stride=(1,))
93
+ )
94
+ (1): Hi_ResidualTemporalBlock(
95
+ (blocks): ModuleList(
96
+ (0-1): 2 x Conv1dBlock(
97
+ (block): Sequential(
98
+ (0): Conv1d(64, 64, kernel_size=(5,), stride=(1,), padding=(2,))
99
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
100
+ (2): GroupNorm(8, 64, eps=1e-05, affine=True)
101
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
102
+ (4): Mish()
103
+ )
104
+ )
105
+ )
106
+ (time_mlp): Sequential(
107
+ (0): Linear(in_features=32, out_features=64, bias=True)
108
+ (1): Mish()
109
+ (2): Linear(in_features=64, out_features=64, bias=True)
110
+ (3): Rearrange('batch t -> batch t 1')
111
+ )
112
+ (residual_conv): Identity()
113
+ )
114
+ (2): Downsample1d(
115
+ (conv): Conv1d(64, 64, kernel_size=(3,), stride=(2,), padding=(1,))
116
+ )
117
+ )
118
+ (2): ModuleList(
119
+ (0): Hi_ResidualTemporalBlock(
120
+ (blocks): ModuleList(
121
+ (0): Conv1dBlock(
122
+ (block): Sequential(
123
+ (0): Conv1d(64, 96, kernel_size=(5,), stride=(1,), padding=(2,))
124
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
125
+ (2): GroupNorm(8, 96, eps=1e-05, affine=True)
126
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
127
+ (4): Mish()
128
+ )
129
+ )
130
+ (1): Conv1dBlock(
131
+ (block): Sequential(
132
+ (0): Conv1d(96, 96, kernel_size=(5,), stride=(1,), padding=(2,))
133
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
134
+ (2): GroupNorm(8, 96, eps=1e-05, affine=True)
135
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
136
+ (4): Mish()
137
+ )
138
+ )
139
+ )
140
+ (time_mlp): Sequential(
141
+ (0): Linear(in_features=32, out_features=64, bias=True)
142
+ (1): Mish()
143
+ (2): Linear(in_features=64, out_features=96, bias=True)
144
+ (3): Rearrange('batch t -> batch t 1')
145
+ )
146
+ (residual_conv): Conv1d(64, 96, kernel_size=(1,), stride=(1,))
147
+ )
148
+ (1): Hi_ResidualTemporalBlock(
149
+ (blocks): ModuleList(
150
+ (0-1): 2 x Conv1dBlock(
151
+ (block): Sequential(
152
+ (0): Conv1d(96, 96, kernel_size=(5,), stride=(1,), padding=(2,))
153
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
154
+ (2): GroupNorm(8, 96, eps=1e-05, affine=True)
155
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
156
+ (4): Mish()
157
+ )
158
+ )
159
+ )
160
+ (time_mlp): Sequential(
161
+ (0): Linear(in_features=32, out_features=64, bias=True)
162
+ (1): Mish()
163
+ (2): Linear(in_features=64, out_features=96, bias=True)
164
+ (3): Rearrange('batch t -> batch t 1')
165
+ )
166
+ (residual_conv): Identity()
167
+ )
168
+ (2): Downsample1d(
169
+ (conv): Conv1d(96, 96, kernel_size=(3,), stride=(2,), padding=(1,))
170
+ )
171
+ )
172
+ (3): ModuleList(
173
+ (0): Hi_ResidualTemporalBlock(
174
+ (blocks): ModuleList(
175
+ (0): Conv1dBlock(
176
+ (block): Sequential(
177
+ (0): Conv1d(96, 128, kernel_size=(5,), stride=(1,), padding=(2,))
178
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
179
+ (2): GroupNorm(8, 128, eps=1e-05, affine=True)
180
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
181
+ (4): Mish()
182
+ )
183
+ )
184
+ (1): Conv1dBlock(
185
+ (block): Sequential(
186
+ (0): Conv1d(128, 128, kernel_size=(5,), stride=(1,), padding=(2,))
187
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
188
+ (2): GroupNorm(8, 128, eps=1e-05, affine=True)
189
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
190
+ (4): Mish()
191
+ )
192
+ )
193
+ )
194
+ (time_mlp): Sequential(
195
+ (0): Linear(in_features=32, out_features=64, bias=True)
196
+ (1): Mish()
197
+ (2): Linear(in_features=64, out_features=128, bias=True)
198
+ (3): Rearrange('batch t -> batch t 1')
199
+ )
200
+ (residual_conv): Conv1d(96, 128, kernel_size=(1,), stride=(1,))
201
+ )
202
+ (1): Hi_ResidualTemporalBlock(
203
+ (blocks): ModuleList(
204
+ (0-1): 2 x Conv1dBlock(
205
+ (block): Sequential(
206
+ (0): Conv1d(128, 128, kernel_size=(5,), stride=(1,), padding=(2,))
207
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
208
+ (2): GroupNorm(8, 128, eps=1e-05, affine=True)
209
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
210
+ (4): Mish()
211
+ )
212
+ )
213
+ )
214
+ (time_mlp): Sequential(
215
+ (0): Linear(in_features=32, out_features=64, bias=True)
216
+ (1): Mish()
217
+ (2): Linear(in_features=64, out_features=128, bias=True)
218
+ (3): Rearrange('batch t -> batch t 1')
219
+ )
220
+ (residual_conv): Identity()
221
+ )
222
+ (2): Downsample1d(
223
+ (conv): Conv1d(128, 128, kernel_size=(3,), stride=(2,), padding=(1,))
224
+ )
225
+ )
226
+ (4): ModuleList(
227
+ (0): Hi_ResidualTemporalBlock(
228
+ (blocks): ModuleList(
229
+ (0): Conv1dBlock(
230
+ (block): Sequential(
231
+ (0): Conv1d(128, 160, kernel_size=(5,), stride=(1,), padding=(2,))
232
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
233
+ (2): GroupNorm(8, 160, eps=1e-05, affine=True)
234
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
235
+ (4): Mish()
236
+ )
237
+ )
238
+ (1): Conv1dBlock(
239
+ (block): Sequential(
240
+ (0): Conv1d(160, 160, kernel_size=(5,), stride=(1,), padding=(2,))
241
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
242
+ (2): GroupNorm(8, 160, eps=1e-05, affine=True)
243
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
244
+ (4): Mish()
245
+ )
246
+ )
247
+ )
248
+ (time_mlp): Sequential(
249
+ (0): Linear(in_features=32, out_features=64, bias=True)
250
+ (1): Mish()
251
+ (2): Linear(in_features=64, out_features=160, bias=True)
252
+ (3): Rearrange('batch t -> batch t 1')
253
+ )
254
+ (residual_conv): Conv1d(128, 160, kernel_size=(1,), stride=(1,))
255
+ )
256
+ (1): Hi_ResidualTemporalBlock(
257
+ (blocks): ModuleList(
258
+ (0-1): 2 x Conv1dBlock(
259
+ (block): Sequential(
260
+ (0): Conv1d(160, 160, kernel_size=(5,), stride=(1,), padding=(2,))
261
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
262
+ (2): GroupNorm(8, 160, eps=1e-05, affine=True)
263
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
264
+ (4): Mish()
265
+ )
266
+ )
267
+ )
268
+ (time_mlp): Sequential(
269
+ (0): Linear(in_features=32, out_features=64, bias=True)
270
+ (1): Mish()
271
+ (2): Linear(in_features=64, out_features=160, bias=True)
272
+ (3): Rearrange('batch t -> batch t 1')
273
+ )
274
+ (residual_conv): Identity()
275
+ )
276
+ (2): Identity()
277
+ )
278
+ )
279
+ (mid_block1): Hi_ResidualTemporalBlock(
280
+ (blocks): ModuleList(
281
+ (0-1): 2 x Conv1dBlock(
282
+ (block): Sequential(
283
+ (0): Conv1d(160, 160, kernel_size=(5,), stride=(1,), padding=(2,))
284
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
285
+ (2): GroupNorm(8, 160, eps=1e-05, affine=True)
286
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
287
+ (4): Mish()
288
+ )
289
+ )
290
+ )
291
+ (time_mlp): Sequential(
292
+ (0): Linear(in_features=32, out_features=64, bias=True)
293
+ (1): Mish()
294
+ (2): Linear(in_features=64, out_features=160, bias=True)
295
+ (3): Rearrange('batch t -> batch t 1')
296
+ )
297
+ (residual_conv): Identity()
298
+ )
299
+ (mid_block2): Hi_ResidualTemporalBlock(
300
+ (blocks): ModuleList(
301
+ (0-1): 2 x Conv1dBlock(
302
+ (block): Sequential(
303
+ (0): Conv1d(160, 160, kernel_size=(5,), stride=(1,), padding=(2,))
304
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
305
+ (2): GroupNorm(8, 160, eps=1e-05, affine=True)
306
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
307
+ (4): Mish()
308
+ )
309
+ )
310
+ )
311
+ (time_mlp): Sequential(
312
+ (0): Linear(in_features=32, out_features=64, bias=True)
313
+ (1): Mish()
314
+ (2): Linear(in_features=64, out_features=160, bias=True)
315
+ (3): Rearrange('batch t -> batch t 1')
316
+ )
317
+ (residual_conv): Identity()
318
+ )
319
+ (final_conv): Sequential(
320
+ (0): Conv1dBlock(
321
+ (block): Sequential(
322
+ (0): Conv1d(160, 160, kernel_size=(3,), stride=(1,), padding=(1,))
323
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
324
+ (2): GroupNorm(8, 160, eps=1e-05, affine=True)
325
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
326
+ (4): Mish()
327
+ )
328
+ )
329
+ (1): Conv1d(160, 160, kernel_size=(1,), stride=(1,))
330
+ )
331
+ (final_mlp): Sequential(
332
+ (0): Linear(in_features=1280, out_features=1360, bias=True)
333
+ (1): Mish()
334
+ (2): Linear(in_features=1360, out_features=512, bias=True)
335
+ (3): Mish()
336
+ (4): Linear(in_features=512, out_features=256, bias=True)
337
+ )
338
+ )
339
+ (end_ovlp_model): Traj_Time_Encoder(
340
+ (time_mlp): Sequential(
341
+ (0): SinusoidalPosEmb()
342
+ (1): Linear(in_features=32, out_features=128, bias=True)
343
+ (2): Mish()
344
+ (3): Linear(in_features=128, out_features=32, bias=True)
345
+ )
346
+ (downs): ModuleList(
347
+ (0): ModuleList(
348
+ (0): Hi_ResidualTemporalBlock(
349
+ (blocks): ModuleList(
350
+ (0): Conv1dBlock(
351
+ (block): Sequential(
352
+ (0): Conv1d(2, 32, kernel_size=(5,), stride=(1,), padding=(2,))
353
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
354
+ (2): GroupNorm(8, 32, eps=1e-05, affine=True)
355
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
356
+ (4): Mish()
357
+ )
358
+ )
359
+ (1): Conv1dBlock(
360
+ (block): Sequential(
361
+ (0): Conv1d(32, 32, kernel_size=(5,), stride=(1,), padding=(2,))
362
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
363
+ (2): GroupNorm(8, 32, eps=1e-05, affine=True)
364
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
365
+ (4): Mish()
366
+ )
367
+ )
368
+ )
369
+ (time_mlp): Sequential(
370
+ (0): Linear(in_features=32, out_features=64, bias=True)
371
+ (1): Mish()
372
+ (2): Linear(in_features=64, out_features=32, bias=True)
373
+ (3): Rearrange('batch t -> batch t 1')
374
+ )
375
+ (residual_conv): Conv1d(2, 32, kernel_size=(1,), stride=(1,))
376
+ )
377
+ (1): Hi_ResidualTemporalBlock(
378
+ (blocks): ModuleList(
379
+ (0-1): 2 x Conv1dBlock(
380
+ (block): Sequential(
381
+ (0): Conv1d(32, 32, kernel_size=(5,), stride=(1,), padding=(2,))
382
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
383
+ (2): GroupNorm(8, 32, eps=1e-05, affine=True)
384
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
385
+ (4): Mish()
386
+ )
387
+ )
388
+ )
389
+ (time_mlp): Sequential(
390
+ (0): Linear(in_features=32, out_features=64, bias=True)
391
+ (1): Mish()
392
+ (2): Linear(in_features=64, out_features=32, bias=True)
393
+ (3): Rearrange('batch t -> batch t 1')
394
+ )
395
+ (residual_conv): Identity()
396
+ )
397
+ (2): Downsample1d(
398
+ (conv): Conv1d(32, 32, kernel_size=(3,), stride=(2,), padding=(1,))
399
+ )
400
+ )
401
+ (1): ModuleList(
402
+ (0): Hi_ResidualTemporalBlock(
403
+ (blocks): ModuleList(
404
+ (0): Conv1dBlock(
405
+ (block): Sequential(
406
+ (0): Conv1d(32, 64, kernel_size=(5,), stride=(1,), padding=(2,))
407
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
408
+ (2): GroupNorm(8, 64, eps=1e-05, affine=True)
409
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
410
+ (4): Mish()
411
+ )
412
+ )
413
+ (1): Conv1dBlock(
414
+ (block): Sequential(
415
+ (0): Conv1d(64, 64, kernel_size=(5,), stride=(1,), padding=(2,))
416
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
417
+ (2): GroupNorm(8, 64, eps=1e-05, affine=True)
418
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
419
+ (4): Mish()
420
+ )
421
+ )
422
+ )
423
+ (time_mlp): Sequential(
424
+ (0): Linear(in_features=32, out_features=64, bias=True)
425
+ (1): Mish()
426
+ (2): Linear(in_features=64, out_features=64, bias=True)
427
+ (3): Rearrange('batch t -> batch t 1')
428
+ )
429
+ (residual_conv): Conv1d(32, 64, kernel_size=(1,), stride=(1,))
430
+ )
431
+ (1): Hi_ResidualTemporalBlock(
432
+ (blocks): ModuleList(
433
+ (0-1): 2 x Conv1dBlock(
434
+ (block): Sequential(
435
+ (0): Conv1d(64, 64, kernel_size=(5,), stride=(1,), padding=(2,))
436
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
437
+ (2): GroupNorm(8, 64, eps=1e-05, affine=True)
438
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
439
+ (4): Mish()
440
+ )
441
+ )
442
+ )
443
+ (time_mlp): Sequential(
444
+ (0): Linear(in_features=32, out_features=64, bias=True)
445
+ (1): Mish()
446
+ (2): Linear(in_features=64, out_features=64, bias=True)
447
+ (3): Rearrange('batch t -> batch t 1')
448
+ )
449
+ (residual_conv): Identity()
450
+ )
451
+ (2): Downsample1d(
452
+ (conv): Conv1d(64, 64, kernel_size=(3,), stride=(2,), padding=(1,))
453
+ )
454
+ )
455
+ (2): ModuleList(
456
+ (0): Hi_ResidualTemporalBlock(
457
+ (blocks): ModuleList(
458
+ (0): Conv1dBlock(
459
+ (block): Sequential(
460
+ (0): Conv1d(64, 96, kernel_size=(5,), stride=(1,), padding=(2,))
461
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
462
+ (2): GroupNorm(8, 96, eps=1e-05, affine=True)
463
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
464
+ (4): Mish()
465
+ )
466
+ )
467
+ (1): Conv1dBlock(
468
+ (block): Sequential(
469
+ (0): Conv1d(96, 96, kernel_size=(5,), stride=(1,), padding=(2,))
470
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
471
+ (2): GroupNorm(8, 96, eps=1e-05, affine=True)
472
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
473
+ (4): Mish()
474
+ )
475
+ )
476
+ )
477
+ (time_mlp): Sequential(
478
+ (0): Linear(in_features=32, out_features=64, bias=True)
479
+ (1): Mish()
480
+ (2): Linear(in_features=64, out_features=96, bias=True)
481
+ (3): Rearrange('batch t -> batch t 1')
482
+ )
483
+ (residual_conv): Conv1d(64, 96, kernel_size=(1,), stride=(1,))
484
+ )
485
+ (1): Hi_ResidualTemporalBlock(
486
+ (blocks): ModuleList(
487
+ (0-1): 2 x Conv1dBlock(
488
+ (block): Sequential(
489
+ (0): Conv1d(96, 96, kernel_size=(5,), stride=(1,), padding=(2,))
490
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
491
+ (2): GroupNorm(8, 96, eps=1e-05, affine=True)
492
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
493
+ (4): Mish()
494
+ )
495
+ )
496
+ )
497
+ (time_mlp): Sequential(
498
+ (0): Linear(in_features=32, out_features=64, bias=True)
499
+ (1): Mish()
500
+ (2): Linear(in_features=64, out_features=96, bias=True)
501
+ (3): Rearrange('batch t -> batch t 1')
502
+ )
503
+ (residual_conv): Identity()
504
+ )
505
+ (2): Downsample1d(
506
+ (conv): Conv1d(96, 96, kernel_size=(3,), stride=(2,), padding=(1,))
507
+ )
508
+ )
509
+ (3): ModuleList(
510
+ (0): Hi_ResidualTemporalBlock(
511
+ (blocks): ModuleList(
512
+ (0): Conv1dBlock(
513
+ (block): Sequential(
514
+ (0): Conv1d(96, 128, kernel_size=(5,), stride=(1,), padding=(2,))
515
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
516
+ (2): GroupNorm(8, 128, eps=1e-05, affine=True)
517
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
518
+ (4): Mish()
519
+ )
520
+ )
521
+ (1): Conv1dBlock(
522
+ (block): Sequential(
523
+ (0): Conv1d(128, 128, kernel_size=(5,), stride=(1,), padding=(2,))
524
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
525
+ (2): GroupNorm(8, 128, eps=1e-05, affine=True)
526
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
527
+ (4): Mish()
528
+ )
529
+ )
530
+ )
531
+ (time_mlp): Sequential(
532
+ (0): Linear(in_features=32, out_features=64, bias=True)
533
+ (1): Mish()
534
+ (2): Linear(in_features=64, out_features=128, bias=True)
535
+ (3): Rearrange('batch t -> batch t 1')
536
+ )
537
+ (residual_conv): Conv1d(96, 128, kernel_size=(1,), stride=(1,))
538
+ )
539
+ (1): Hi_ResidualTemporalBlock(
540
+ (blocks): ModuleList(
541
+ (0-1): 2 x Conv1dBlock(
542
+ (block): Sequential(
543
+ (0): Conv1d(128, 128, kernel_size=(5,), stride=(1,), padding=(2,))
544
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
545
+ (2): GroupNorm(8, 128, eps=1e-05, affine=True)
546
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
547
+ (4): Mish()
548
+ )
549
+ )
550
+ )
551
+ (time_mlp): Sequential(
552
+ (0): Linear(in_features=32, out_features=64, bias=True)
553
+ (1): Mish()
554
+ (2): Linear(in_features=64, out_features=128, bias=True)
555
+ (3): Rearrange('batch t -> batch t 1')
556
+ )
557
+ (residual_conv): Identity()
558
+ )
559
+ (2): Downsample1d(
560
+ (conv): Conv1d(128, 128, kernel_size=(3,), stride=(2,), padding=(1,))
561
+ )
562
+ )
563
+ (4): ModuleList(
564
+ (0): Hi_ResidualTemporalBlock(
565
+ (blocks): ModuleList(
566
+ (0): Conv1dBlock(
567
+ (block): Sequential(
568
+ (0): Conv1d(128, 160, kernel_size=(5,), stride=(1,), padding=(2,))
569
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
570
+ (2): GroupNorm(8, 160, eps=1e-05, affine=True)
571
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
572
+ (4): Mish()
573
+ )
574
+ )
575
+ (1): Conv1dBlock(
576
+ (block): Sequential(
577
+ (0): Conv1d(160, 160, kernel_size=(5,), stride=(1,), padding=(2,))
578
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
579
+ (2): GroupNorm(8, 160, eps=1e-05, affine=True)
580
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
581
+ (4): Mish()
582
+ )
583
+ )
584
+ )
585
+ (time_mlp): Sequential(
586
+ (0): Linear(in_features=32, out_features=64, bias=True)
587
+ (1): Mish()
588
+ (2): Linear(in_features=64, out_features=160, bias=True)
589
+ (3): Rearrange('batch t -> batch t 1')
590
+ )
591
+ (residual_conv): Conv1d(128, 160, kernel_size=(1,), stride=(1,))
592
+ )
593
+ (1): Hi_ResidualTemporalBlock(
594
+ (blocks): ModuleList(
595
+ (0-1): 2 x Conv1dBlock(
596
+ (block): Sequential(
597
+ (0): Conv1d(160, 160, kernel_size=(5,), stride=(1,), padding=(2,))
598
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
599
+ (2): GroupNorm(8, 160, eps=1e-05, affine=True)
600
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
601
+ (4): Mish()
602
+ )
603
+ )
604
+ )
605
+ (time_mlp): Sequential(
606
+ (0): Linear(in_features=32, out_features=64, bias=True)
607
+ (1): Mish()
608
+ (2): Linear(in_features=64, out_features=160, bias=True)
609
+ (3): Rearrange('batch t -> batch t 1')
610
+ )
611
+ (residual_conv): Identity()
612
+ )
613
+ (2): Identity()
614
+ )
615
+ )
616
+ (mid_block1): Hi_ResidualTemporalBlock(
617
+ (blocks): ModuleList(
618
+ (0-1): 2 x Conv1dBlock(
619
+ (block): Sequential(
620
+ (0): Conv1d(160, 160, kernel_size=(5,), stride=(1,), padding=(2,))
621
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
622
+ (2): GroupNorm(8, 160, eps=1e-05, affine=True)
623
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
624
+ (4): Mish()
625
+ )
626
+ )
627
+ )
628
+ (time_mlp): Sequential(
629
+ (0): Linear(in_features=32, out_features=64, bias=True)
630
+ (1): Mish()
631
+ (2): Linear(in_features=64, out_features=160, bias=True)
632
+ (3): Rearrange('batch t -> batch t 1')
633
+ )
634
+ (residual_conv): Identity()
635
+ )
636
+ (mid_block2): Hi_ResidualTemporalBlock(
637
+ (blocks): ModuleList(
638
+ (0-1): 2 x Conv1dBlock(
639
+ (block): Sequential(
640
+ (0): Conv1d(160, 160, kernel_size=(5,), stride=(1,), padding=(2,))
641
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
642
+ (2): GroupNorm(8, 160, eps=1e-05, affine=True)
643
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
644
+ (4): Mish()
645
+ )
646
+ )
647
+ )
648
+ (time_mlp): Sequential(
649
+ (0): Linear(in_features=32, out_features=64, bias=True)
650
+ (1): Mish()
651
+ (2): Linear(in_features=64, out_features=160, bias=True)
652
+ (3): Rearrange('batch t -> batch t 1')
653
+ )
654
+ (residual_conv): Identity()
655
+ )
656
+ (final_conv): Sequential(
657
+ (0): Conv1dBlock(
658
+ (block): Sequential(
659
+ (0): Conv1d(160, 160, kernel_size=(3,), stride=(1,), padding=(1,))
660
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
661
+ (2): GroupNorm(8, 160, eps=1e-05, affine=True)
662
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
663
+ (4): Mish()
664
+ )
665
+ )
666
+ (1): Conv1d(160, 160, kernel_size=(1,), stride=(1,))
667
+ )
668
+ (final_mlp): Sequential(
669
+ (0): Linear(in_features=1280, out_features=1360, bias=True)
670
+ (1): Mish()
671
+ (2): Linear(in_features=1360, out_features=512, bias=True)
672
+ (3): Mish()
673
+ (4): Linear(in_features=512, out_features=256, bias=True)
674
+ )
675
+ )
676
+ (st_inpaint_model): Identity()
677
+ (end_inpaint_model): Identity()
678
+ (time_mlp): Sequential(
679
+ (0): SinusoidalPosEmb()
680
+ (1): Linear(in_features=96, out_features=192, bias=True)
681
+ (2): Mish()
682
+ (3): Linear(in_features=192, out_features=192, bias=True)
683
+ (4): Mish()
684
+ (5): Linear(in_features=192, out_features=96, bias=True)
685
+ )
686
+ (downs): ModuleList(
687
+ (0): ModuleList(
688
+ (0): ResidualTemporalBlock_dd(
689
+ (blocks): ModuleList(
690
+ (0): Conv1dBlock_dd(
691
+ (block): Sequential(
692
+ (0): Conv1d(2, 128, kernel_size=(5,), stride=(1,), padding=(2,))
693
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
694
+ (2): GroupNorm(8, 128, eps=1e-05, affine=True)
695
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
696
+ (4): Mish()
697
+ )
698
+ )
699
+ (1): Conv1dBlock_dd(
700
+ (block): Sequential(
701
+ (0): Conv1d(128, 128, kernel_size=(5,), stride=(1,), padding=(2,))
702
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
703
+ (2): GroupNorm(8, 128, eps=1e-05, affine=True)
704
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
705
+ (4): Mish()
706
+ )
707
+ )
708
+ )
709
+ (time_mlp): Sequential(
710
+ (0): Linear(in_features=672, out_features=1344, bias=True)
711
+ (1): Mish()
712
+ (2): Linear(in_features=1344, out_features=128, bias=True)
713
+ (3): Rearrange('batch t -> batch t 1')
714
+ )
715
+ (residual_conv): Conv1d(2, 128, kernel_size=(1,), stride=(1,))
716
+ )
717
+ (1): ResidualTemporalBlock_dd(
718
+ (blocks): ModuleList(
719
+ (0-1): 2 x Conv1dBlock_dd(
720
+ (block): Sequential(
721
+ (0): Conv1d(128, 128, kernel_size=(5,), stride=(1,), padding=(2,))
722
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
723
+ (2): GroupNorm(8, 128, eps=1e-05, affine=True)
724
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
725
+ (4): Mish()
726
+ )
727
+ )
728
+ )
729
+ (time_mlp): Sequential(
730
+ (0): Linear(in_features=672, out_features=1344, bias=True)
731
+ (1): Mish()
732
+ (2): Linear(in_features=1344, out_features=128, bias=True)
733
+ (3): Rearrange('batch t -> batch t 1')
734
+ )
735
+ (residual_conv): Identity()
736
+ )
737
+ (2): Downsample1d(
738
+ (conv): Conv1d(128, 128, kernel_size=(3,), stride=(2,), padding=(1,))
739
+ )
740
+ )
741
+ (1): ModuleList(
742
+ (0): ResidualTemporalBlock_dd(
743
+ (blocks): ModuleList(
744
+ (0): Conv1dBlock_dd(
745
+ (block): Sequential(
746
+ (0): Conv1d(128, 256, kernel_size=(5,), stride=(1,), padding=(2,))
747
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
748
+ (2): GroupNorm(8, 256, eps=1e-05, affine=True)
749
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
750
+ (4): Mish()
751
+ )
752
+ )
753
+ (1): Conv1dBlock_dd(
754
+ (block): Sequential(
755
+ (0): Conv1d(256, 256, kernel_size=(5,), stride=(1,), padding=(2,))
756
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
757
+ (2): GroupNorm(8, 256, eps=1e-05, affine=True)
758
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
759
+ (4): Mish()
760
+ )
761
+ )
762
+ )
763
+ (time_mlp): Sequential(
764
+ (0): Linear(in_features=672, out_features=1344, bias=True)
765
+ (1): Mish()
766
+ (2): Linear(in_features=1344, out_features=256, bias=True)
767
+ (3): Rearrange('batch t -> batch t 1')
768
+ )
769
+ (residual_conv): Conv1d(128, 256, kernel_size=(1,), stride=(1,))
770
+ )
771
+ (1): ResidualTemporalBlock_dd(
772
+ (blocks): ModuleList(
773
+ (0-1): 2 x Conv1dBlock_dd(
774
+ (block): Sequential(
775
+ (0): Conv1d(256, 256, kernel_size=(5,), stride=(1,), padding=(2,))
776
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
777
+ (2): GroupNorm(8, 256, eps=1e-05, affine=True)
778
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
779
+ (4): Mish()
780
+ )
781
+ )
782
+ )
783
+ (time_mlp): Sequential(
784
+ (0): Linear(in_features=672, out_features=1344, bias=True)
785
+ (1): Mish()
786
+ (2): Linear(in_features=1344, out_features=256, bias=True)
787
+ (3): Rearrange('batch t -> batch t 1')
788
+ )
789
+ (residual_conv): Identity()
790
+ )
791
+ (2): Downsample1d(
792
+ (conv): Conv1d(256, 256, kernel_size=(3,), stride=(2,), padding=(1,))
793
+ )
794
+ )
795
+ (2): ModuleList(
796
+ (0): ResidualTemporalBlock_dd(
797
+ (blocks): ModuleList(
798
+ (0): Conv1dBlock_dd(
799
+ (block): Sequential(
800
+ (0): Conv1d(256, 512, kernel_size=(5,), stride=(1,), padding=(2,))
801
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
802
+ (2): GroupNorm(8, 512, eps=1e-05, affine=True)
803
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
804
+ (4): Mish()
805
+ )
806
+ )
807
+ (1): Conv1dBlock_dd(
808
+ (block): Sequential(
809
+ (0): Conv1d(512, 512, kernel_size=(5,), stride=(1,), padding=(2,))
810
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
811
+ (2): GroupNorm(8, 512, eps=1e-05, affine=True)
812
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
813
+ (4): Mish()
814
+ )
815
+ )
816
+ )
817
+ (time_mlp): Sequential(
818
+ (0): Linear(in_features=672, out_features=1344, bias=True)
819
+ (1): Mish()
820
+ (2): Linear(in_features=1344, out_features=512, bias=True)
821
+ (3): Rearrange('batch t -> batch t 1')
822
+ )
823
+ (residual_conv): Conv1d(256, 512, kernel_size=(1,), stride=(1,))
824
+ )
825
+ (1): ResidualTemporalBlock_dd(
826
+ (blocks): ModuleList(
827
+ (0-1): 2 x Conv1dBlock_dd(
828
+ (block): Sequential(
829
+ (0): Conv1d(512, 512, kernel_size=(5,), stride=(1,), padding=(2,))
830
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
831
+ (2): GroupNorm(8, 512, eps=1e-05, affine=True)
832
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
833
+ (4): Mish()
834
+ )
835
+ )
836
+ )
837
+ (time_mlp): Sequential(
838
+ (0): Linear(in_features=672, out_features=1344, bias=True)
839
+ (1): Mish()
840
+ (2): Linear(in_features=1344, out_features=512, bias=True)
841
+ (3): Rearrange('batch t -> batch t 1')
842
+ )
843
+ (residual_conv): Identity()
844
+ )
845
+ (2): Downsample1d(
846
+ (conv): Conv1d(512, 512, kernel_size=(3,), stride=(2,), padding=(1,))
847
+ )
848
+ )
849
+ (3): ModuleList(
850
+ (0): ResidualTemporalBlock_dd(
851
+ (blocks): ModuleList(
852
+ (0): Conv1dBlock_dd(
853
+ (block): Sequential(
854
+ (0): Conv1d(512, 1024, kernel_size=(5,), stride=(1,), padding=(2,))
855
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
856
+ (2): GroupNorm(8, 1024, eps=1e-05, affine=True)
857
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
858
+ (4): Mish()
859
+ )
860
+ )
861
+ (1): Conv1dBlock_dd(
862
+ (block): Sequential(
863
+ (0): Conv1d(1024, 1024, kernel_size=(5,), stride=(1,), padding=(2,))
864
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
865
+ (2): GroupNorm(8, 1024, eps=1e-05, affine=True)
866
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
867
+ (4): Mish()
868
+ )
869
+ )
870
+ )
871
+ (time_mlp): Sequential(
872
+ (0): Linear(in_features=672, out_features=1344, bias=True)
873
+ (1): Mish()
874
+ (2): Linear(in_features=1344, out_features=1024, bias=True)
875
+ (3): Rearrange('batch t -> batch t 1')
876
+ )
877
+ (residual_conv): Conv1d(512, 1024, kernel_size=(1,), stride=(1,))
878
+ )
879
+ (1): ResidualTemporalBlock_dd(
880
+ (blocks): ModuleList(
881
+ (0-1): 2 x Conv1dBlock_dd(
882
+ (block): Sequential(
883
+ (0): Conv1d(1024, 1024, kernel_size=(5,), stride=(1,), padding=(2,))
884
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
885
+ (2): GroupNorm(8, 1024, eps=1e-05, affine=True)
886
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
887
+ (4): Mish()
888
+ )
889
+ )
890
+ )
891
+ (time_mlp): Sequential(
892
+ (0): Linear(in_features=672, out_features=1344, bias=True)
893
+ (1): Mish()
894
+ (2): Linear(in_features=1344, out_features=1024, bias=True)
895
+ (3): Rearrange('batch t -> batch t 1')
896
+ )
897
+ (residual_conv): Identity()
898
+ )
899
+ (2): Identity()
900
+ )
901
+ )
902
+ (ups): ModuleList(
903
+ (0): ModuleList(
904
+ (0): ResidualTemporalBlock_dd(
905
+ (blocks): ModuleList(
906
+ (0): Conv1dBlock_dd(
907
+ (block): Sequential(
908
+ (0): Conv1d(2048, 512, kernel_size=(5,), stride=(1,), padding=(2,))
909
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
910
+ (2): GroupNorm(8, 512, eps=1e-05, affine=True)
911
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
912
+ (4): Mish()
913
+ )
914
+ )
915
+ (1): Conv1dBlock_dd(
916
+ (block): Sequential(
917
+ (0): Conv1d(512, 512, kernel_size=(5,), stride=(1,), padding=(2,))
918
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
919
+ (2): GroupNorm(8, 512, eps=1e-05, affine=True)
920
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
921
+ (4): Mish()
922
+ )
923
+ )
924
+ )
925
+ (time_mlp): Sequential(
926
+ (0): Linear(in_features=672, out_features=1344, bias=True)
927
+ (1): Mish()
928
+ (2): Linear(in_features=1344, out_features=512, bias=True)
929
+ (3): Rearrange('batch t -> batch t 1')
930
+ )
931
+ (residual_conv): Conv1d(2048, 512, kernel_size=(1,), stride=(1,))
932
+ )
933
+ (1): ResidualTemporalBlock_dd(
934
+ (blocks): ModuleList(
935
+ (0-1): 2 x Conv1dBlock_dd(
936
+ (block): Sequential(
937
+ (0): Conv1d(512, 512, kernel_size=(5,), stride=(1,), padding=(2,))
938
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
939
+ (2): GroupNorm(8, 512, eps=1e-05, affine=True)
940
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
941
+ (4): Mish()
942
+ )
943
+ )
944
+ )
945
+ (time_mlp): Sequential(
946
+ (0): Linear(in_features=672, out_features=1344, bias=True)
947
+ (1): Mish()
948
+ (2): Linear(in_features=1344, out_features=512, bias=True)
949
+ (3): Rearrange('batch t -> batch t 1')
950
+ )
951
+ (residual_conv): Identity()
952
+ )
953
+ (2): Upsample1d(
954
+ (conv): ConvTranspose1d(512, 512, kernel_size=(4,), stride=(2,), padding=(1,))
955
+ )
956
+ )
957
+ (1): ModuleList(
958
+ (0): ResidualTemporalBlock_dd(
959
+ (blocks): ModuleList(
960
+ (0): Conv1dBlock_dd(
961
+ (block): Sequential(
962
+ (0): Conv1d(1024, 256, kernel_size=(5,), stride=(1,), padding=(2,))
963
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
964
+ (2): GroupNorm(8, 256, eps=1e-05, affine=True)
965
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
966
+ (4): Mish()
967
+ )
968
+ )
969
+ (1): Conv1dBlock_dd(
970
+ (block): Sequential(
971
+ (0): Conv1d(256, 256, kernel_size=(5,), stride=(1,), padding=(2,))
972
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
973
+ (2): GroupNorm(8, 256, eps=1e-05, affine=True)
974
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
975
+ (4): Mish()
976
+ )
977
+ )
978
+ )
979
+ (time_mlp): Sequential(
980
+ (0): Linear(in_features=672, out_features=1344, bias=True)
981
+ (1): Mish()
982
+ (2): Linear(in_features=1344, out_features=256, bias=True)
983
+ (3): Rearrange('batch t -> batch t 1')
984
+ )
985
+ (residual_conv): Conv1d(1024, 256, kernel_size=(1,), stride=(1,))
986
+ )
987
+ (1): ResidualTemporalBlock_dd(
988
+ (blocks): ModuleList(
989
+ (0-1): 2 x Conv1dBlock_dd(
990
+ (block): Sequential(
991
+ (0): Conv1d(256, 256, kernel_size=(5,), stride=(1,), padding=(2,))
992
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
993
+ (2): GroupNorm(8, 256, eps=1e-05, affine=True)
994
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
995
+ (4): Mish()
996
+ )
997
+ )
998
+ )
999
+ (time_mlp): Sequential(
1000
+ (0): Linear(in_features=672, out_features=1344, bias=True)
1001
+ (1): Mish()
1002
+ (2): Linear(in_features=1344, out_features=256, bias=True)
1003
+ (3): Rearrange('batch t -> batch t 1')
1004
+ )
1005
+ (residual_conv): Identity()
1006
+ )
1007
+ (2): Upsample1d(
1008
+ (conv): ConvTranspose1d(256, 256, kernel_size=(4,), stride=(2,), padding=(1,))
1009
+ )
1010
+ )
1011
+ (2): ModuleList(
1012
+ (0): ResidualTemporalBlock_dd(
1013
+ (blocks): ModuleList(
1014
+ (0): Conv1dBlock_dd(
1015
+ (block): Sequential(
1016
+ (0): Conv1d(512, 128, kernel_size=(5,), stride=(1,), padding=(2,))
1017
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
1018
+ (2): GroupNorm(8, 128, eps=1e-05, affine=True)
1019
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
1020
+ (4): Mish()
1021
+ )
1022
+ )
1023
+ (1): Conv1dBlock_dd(
1024
+ (block): Sequential(
1025
+ (0): Conv1d(128, 128, kernel_size=(5,), stride=(1,), padding=(2,))
1026
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
1027
+ (2): GroupNorm(8, 128, eps=1e-05, affine=True)
1028
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
1029
+ (4): Mish()
1030
+ )
1031
+ )
1032
+ )
1033
+ (time_mlp): Sequential(
1034
+ (0): Linear(in_features=672, out_features=1344, bias=True)
1035
+ (1): Mish()
1036
+ (2): Linear(in_features=1344, out_features=128, bias=True)
1037
+ (3): Rearrange('batch t -> batch t 1')
1038
+ )
1039
+ (residual_conv): Conv1d(512, 128, kernel_size=(1,), stride=(1,))
1040
+ )
1041
+ (1): ResidualTemporalBlock_dd(
1042
+ (blocks): ModuleList(
1043
+ (0-1): 2 x Conv1dBlock_dd(
1044
+ (block): Sequential(
1045
+ (0): Conv1d(128, 128, kernel_size=(5,), stride=(1,), padding=(2,))
1046
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
1047
+ (2): GroupNorm(8, 128, eps=1e-05, affine=True)
1048
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
1049
+ (4): Mish()
1050
+ )
1051
+ )
1052
+ )
1053
+ (time_mlp): Sequential(
1054
+ (0): Linear(in_features=672, out_features=1344, bias=True)
1055
+ (1): Mish()
1056
+ (2): Linear(in_features=1344, out_features=128, bias=True)
1057
+ (3): Rearrange('batch t -> batch t 1')
1058
+ )
1059
+ (residual_conv): Identity()
1060
+ )
1061
+ (2): Upsample1d(
1062
+ (conv): ConvTranspose1d(128, 128, kernel_size=(4,), stride=(2,), padding=(1,))
1063
+ )
1064
+ )
1065
+ )
1066
+ (mid_block1): ResidualTemporalBlock_dd(
1067
+ (blocks): ModuleList(
1068
+ (0-1): 2 x Conv1dBlock_dd(
1069
+ (block): Sequential(
1070
+ (0): Conv1d(1024, 1024, kernel_size=(5,), stride=(1,), padding=(2,))
1071
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
1072
+ (2): GroupNorm(8, 1024, eps=1e-05, affine=True)
1073
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
1074
+ (4): Mish()
1075
+ )
1076
+ )
1077
+ )
1078
+ (time_mlp): Sequential(
1079
+ (0): Linear(in_features=672, out_features=1344, bias=True)
1080
+ (1): Mish()
1081
+ (2): Linear(in_features=1344, out_features=1024, bias=True)
1082
+ (3): Rearrange('batch t -> batch t 1')
1083
+ )
1084
+ (residual_conv): Identity()
1085
+ )
1086
+ (mid_block2): ResidualTemporalBlock_dd(
1087
+ (blocks): ModuleList(
1088
+ (0-1): 2 x Conv1dBlock_dd(
1089
+ (block): Sequential(
1090
+ (0): Conv1d(1024, 1024, kernel_size=(5,), stride=(1,), padding=(2,))
1091
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
1092
+ (2): GroupNorm(8, 1024, eps=1e-05, affine=True)
1093
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
1094
+ (4): Mish()
1095
+ )
1096
+ )
1097
+ )
1098
+ (time_mlp): Sequential(
1099
+ (0): Linear(in_features=672, out_features=1344, bias=True)
1100
+ (1): Mish()
1101
+ (2): Linear(in_features=1344, out_features=1024, bias=True)
1102
+ (3): Rearrange('batch t -> batch t 1')
1103
+ )
1104
+ (residual_conv): Identity()
1105
+ )
1106
+ (final_conv): Sequential(
1107
+ (0): Conv1dBlock(
1108
+ (block): Sequential(
1109
+ (0): Conv1d(128, 128, kernel_size=(5,), stride=(1,), padding=(2,))
1110
+ (1): Rearrange('batch channels horizon -> batch channels 1 horizon')
1111
+ (2): GroupNorm(8, 128, eps=1e-05, affine=True)
1112
+ (3): Rearrange('batch channels 1 horizon -> batch channels horizon')
1113
+ (4): Mish()
1114
+ )
1115
+ )
1116
+ (1): Conv1d(128, 2, kernel_size=(1,), stride=(1,))
1117
+ )
1118
+ )