valentinospariza commited on
Commit
db937aa
·
verified ·
1 Parent(s): bc1ef4f

Add training metadata and checksums: geometry_ratio/depth10_normals20

Browse files
ablation_models/additional_studies/geometry/depth_normal_allocation/depth10_normals20/checkpoint-37500/COMPLETE.json ADDED
@@ -0,0 +1,7 @@
 
 
 
 
 
 
 
 
1
+ {
2
+ "status": "COMPLETE",
3
+ "attempt": "1",
4
+ "completed_utc": "2026-08-12T09:42:36.751883+00:00",
5
+ "completion_source": "trainer_exit_zero",
6
+ "wrapper_exit_code": 0
7
+ }
ablation_models/additional_studies/geometry/depth_normal_allocation/depth10_normals20/checkpoint-37500/SHA256SUMS ADDED
@@ -0,0 +1,6 @@
 
 
 
 
 
 
 
1
+ be895c2602b3d07fbbe7abb8ef7322315c4d5aab5db822a4fdc5f56ee91db42a config.json
2
+ 235a30e4d3ac93b6eaf483fe690af3d302d18b9e883878695fa60aff5ff0818b diffusion_pytorch_model.safetensors.index.json
3
+ 242f5575ae52c98692cf6b7b5a8eb9b47e4064d9476d52e3a59f645105ae1775 diffusion_pytorch_model-00001-of-00004.safetensors
4
+ 020d5bbd08c2c7bf0c16ed140f1a45da0c17cb50383a4a241d0b0d4e631d5e44 diffusion_pytorch_model-00002-of-00004.safetensors
5
+ f24a7331cd8fea4c5c9fa1a3f93c2159266e40935cd46186a469bc6b17e79467 diffusion_pytorch_model-00003-of-00004.safetensors
6
+ 772b7a06e1b5cf85b6027a54a5439b26284cc1149aa829c72bb02e07c3c2de57 diffusion_pytorch_model-00004-of-00004.safetensors
ablation_models/additional_studies/geometry/depth_normal_allocation/depth10_normals20/checkpoint-37500/checksum_validation.json ADDED
@@ -0,0 +1,30 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "fingerprints": {
3
+ "config.json": {
4
+ "size": 563,
5
+ "mtime_ns": 1786527721000000000
6
+ },
7
+ "diffusion_pytorch_model.safetensors.index.json": {
8
+ "size": 24875,
9
+ "mtime_ns": 1786527750000000000
10
+ },
11
+ "diffusion_pytorch_model-00001-of-00004.safetensors": {
12
+ "size": 9804206048,
13
+ "mtime_ns": 1786527729000000000
14
+ },
15
+ "diffusion_pytorch_model-00002-of-00004.safetensors": {
16
+ "size": 9797913704,
17
+ "mtime_ns": 1786527736000000000
18
+ },
19
+ "diffusion_pytorch_model-00003-of-00004.safetensors": {
20
+ "size": 9596584288,
21
+ "mtime_ns": 1786527744000000000
22
+ },
23
+ "diffusion_pytorch_model-00004-of-00004.safetensors": {
24
+ "size": 7115649136,
25
+ "mtime_ns": 1786527750000000000
26
+ }
27
+ },
28
+ "sha256_manifest": "6f638635a6c80d0cb359791c0b27ca1ea6f9d6a01832c4eb991933586bb1a1ff",
29
+ "validation": "PASS"
30
+ }
ablation_models/additional_studies/geometry/depth_normal_allocation/depth10_normals20/checkpoint-37500/dataset_specs.json ADDED
@@ -0,0 +1,282 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "datasets": [
3
+ {
4
+ "filters": {
5
+ "source_dataset": "acdc"
6
+ },
7
+ "index_path": "<local-path>",
8
+ "name": "semseg_acdc",
9
+ "path": "<local-path>",
10
+ "sampling": "row",
11
+ "type": "indexed_prepared_tar",
12
+ "weight": 0.035
13
+ },
14
+ {
15
+ "filters": {
16
+ "source_dataset": "bdd"
17
+ },
18
+ "index_path": "<local-path>",
19
+ "name": "semseg_bdd",
20
+ "path": "<local-path>",
21
+ "sampling": "row",
22
+ "type": "indexed_prepared_tar",
23
+ "weight": 0.035
24
+ },
25
+ {
26
+ "filters": {
27
+ "source_dataset": "ade20k"
28
+ },
29
+ "index_path": "<local-path>",
30
+ "name": "semseg_ade20k",
31
+ "path": "<local-path>",
32
+ "sampling": "row",
33
+ "type": "indexed_prepared_tar",
34
+ "weight": 0.035
35
+ },
36
+ {
37
+ "filters": {
38
+ "source_dataset": "voc"
39
+ },
40
+ "index_path": "<local-path>",
41
+ "name": "semseg_voc",
42
+ "path": "<local-path>",
43
+ "sampling": "row",
44
+ "type": "indexed_prepared_tar",
45
+ "weight": 0.035
46
+ },
47
+ {
48
+ "filters": {
49
+ "source_dataset": "ade20k_inst"
50
+ },
51
+ "index_path": "<local-path>",
52
+ "name": "instanceseg_ade20k",
53
+ "path": "<local-path>",
54
+ "sampling": "row",
55
+ "type": "indexed_prepared_tar",
56
+ "weight": 0.032
57
+ },
58
+ {
59
+ "filters": {
60
+ "source_dataset": "mapillary"
61
+ },
62
+ "index_path": "<local-path>",
63
+ "name": "instanceseg_mapillary",
64
+ "path": "<local-path>",
65
+ "sampling": "row",
66
+ "type": "indexed_prepared_tar",
67
+ "weight": 0.032
68
+ },
69
+ {
70
+ "filters": {
71
+ "source_dataset": "voc_inst"
72
+ },
73
+ "index_path": "<local-path>",
74
+ "name": "instanceseg_voc",
75
+ "path": "<local-path>",
76
+ "sampling": "row",
77
+ "type": "indexed_prepared_tar",
78
+ "weight": 0.016
79
+ },
80
+ {
81
+ "index_path": "<local-path>",
82
+ "name": "refseg",
83
+ "path": "<local-path>",
84
+ "sampling": "row",
85
+ "type": "indexed_prepared_tar",
86
+ "weight": 0.08
87
+ },
88
+ {
89
+ "index_path": "<local-path>",
90
+ "name": "pico_banana",
91
+ "path": "<local-path>",
92
+ "sampling": "row",
93
+ "type": "indexed_prepared_tar",
94
+ "weight": 0.15
95
+ },
96
+ {
97
+ "index_path": "<local-path>",
98
+ "name": "identity",
99
+ "path": "<local-path>",
100
+ "sampling": "row",
101
+ "type": "indexed_prepared_tar",
102
+ "weight": 0.1
103
+ },
104
+ {
105
+ "name": "pd12m",
106
+ "shards": "<local-path>",
107
+ "source_dataset": "pd12m",
108
+ "streaming": {
109
+ "error_policy": "warn_and_continue",
110
+ "resampled": true,
111
+ "sample_shuffle_buffer": 10000,
112
+ "shard_shuffle": true
113
+ },
114
+ "task": "text_to_image",
115
+ "type": "webdataset_text_to_image",
116
+ "weight": 0.15
117
+ },
118
+ {
119
+ "archive_sha256": "36aac18f3ff4a51226def096a8f9ffd3bb4ee8f86ccda9e78f19ba50c407c344",
120
+ "geometry_version": "geometry_canonical_v2",
121
+ "index_path": "<local-path>",
122
+ "index_sha256": "ceed5191233ce3b4bf1dd14fa869ca0f935bd0604f9a3ffd9d6f1d3ff326d632",
123
+ "name": "depth_hypersim",
124
+ "path": "<local-path>",
125
+ "sample_list_sha256": "f7b774dbf2645993955fbd315e2a0edc7f499ba3a9513a4101f481dc7a1a9cf7",
126
+ "sampling": "scene_balanced",
127
+ "source_dataset": "hypersim",
128
+ "task": "depth_estimation",
129
+ "type": "indexed_prepared_tar",
130
+ "validation_report": "<local-path>",
131
+ "weight": 0.02
132
+ },
133
+ {
134
+ "archive_sha256": "8968b57be59dc69a66d5cffaff7667e6836188e7c849003bdf66e5e4ff01b18a",
135
+ "geometry_version": "geometry_canonical_v2",
136
+ "index_path": "<local-path>",
137
+ "index_sha256": "ac4f12e2ddcb33ee53425ccc085d4c2ac07a2c776e0e9b20f3bbf371ae636e55",
138
+ "name": "depth_matterport3d",
139
+ "path": "<local-path>",
140
+ "sample_list_sha256": "37280ca662f240992afc5fc0400f1a5d27d4bbb08f61665bd46ac3d27cb0e5ff",
141
+ "sampling": "scene_balanced",
142
+ "source_dataset": "matterport3d",
143
+ "task": "depth_estimation",
144
+ "type": "indexed_prepared_tar",
145
+ "validation_report": "<local-path>",
146
+ "weight": 0.02
147
+ },
148
+ {
149
+ "archive_sha256": "0840146fd580149b35cf67207a176812c207851a0880b0dbfbe5b479c1417246",
150
+ "geometry_version": "geometry_canonical_v2",
151
+ "index_path": "<local-path>",
152
+ "index_sha256": "327b3e2cbec5092f6606ee551f9387b524f5493693ed1bc9ee4353ceca1268d9",
153
+ "name": "depth_replica",
154
+ "path": "<local-path>",
155
+ "sample_list_sha256": "89bcff6ca51ff82624ff44aa986bf6982fe7863195f63d4865e889c8560700bb",
156
+ "sampling": "scene_balanced",
157
+ "source_dataset": "replica",
158
+ "task": "depth_estimation",
159
+ "type": "indexed_prepared_tar",
160
+ "validation_report": "<local-path>",
161
+ "weight": 0.02
162
+ },
163
+ {
164
+ "archive_sha256": "7d62865ca5768e5888b88d9fab9e9de9d6a44cf796e3a22d86105b6c3d0e1d33",
165
+ "geometry_version": "geometry_canonical_v2",
166
+ "index_path": "<local-path>",
167
+ "index_sha256": "381e6a1454f6aceae4030ca2bbb23a6f855e1608a3bc6213b2b14c675c3c3f7f",
168
+ "name": "depth_synscapes",
169
+ "path": "<local-path>",
170
+ "sample_list_sha256": "7118294ad359f2cb9c1acb6a93719a3a6769f3dbaa4b646a216f8788de961520",
171
+ "sampling": "row",
172
+ "source_dataset": "synscapes",
173
+ "task": "depth_estimation",
174
+ "type": "indexed_prepared_tar",
175
+ "validation_report": "<local-path>",
176
+ "weight": 0.02
177
+ },
178
+ {
179
+ "archive_sha256": "179cec3d7309089a9c16c41a6ec2a8f6339d30e96feb633453f113793507963b",
180
+ "geometry_version": "geometry_canonical_v2",
181
+ "index_path": "<local-path>",
182
+ "index_sha256": "2821d90b4432419169d7f98cbd515bffc2069cc6016a317c2c23c260d0e2eaff",
183
+ "name": "depth_taskonomy",
184
+ "path": "<local-path>",
185
+ "sample_list_sha256": "698aad7a87803c74ff750bfaa2cd37fbdf29d4b5b73e77009aea4b41c424c7d2",
186
+ "sampling": "scene_balanced",
187
+ "source_dataset": "taskonomy",
188
+ "task": "depth_estimation",
189
+ "type": "indexed_prepared_tar",
190
+ "validation_report": "<local-path>",
191
+ "weight": 0.02
192
+ },
193
+ {
194
+ "archive_sha256": "40a0424cfb4484ea5edf7ca773ecbc52d5cb603240a4bc6e12afa085d0db0fba",
195
+ "geometry_version": "geometry_mixed_v2_v3",
196
+ "index_path": "<local-path>",
197
+ "index_sha256": "e4346e4a727ba3e1cdc58391368c51025e242d54b2b8f69cc795162e1a3e347c",
198
+ "name": "normals_hypersim",
199
+ "path": "<local-path>",
200
+ "sample_list_sha256": "34d6036cea27350c8cbba0a51614388561ce08b89871eafd0094c3b758572fd0",
201
+ "sampling": "scene_balanced",
202
+ "source_dataset": "hypersim",
203
+ "task": "surface_normal_estimation",
204
+ "type": "indexed_prepared_tar",
205
+ "validation_report": "<local-path>",
206
+ "weight": 0.04
207
+ },
208
+ {
209
+ "archive_sha256": "735630652f130fd7209b5f53ca5343d141935e15166fdf939db14f1996b891d5",
210
+ "geometry_version": "geometry_canonical_v2",
211
+ "index_path": "<local-path>",
212
+ "index_sha256": "1c19051fdd1c248ce986f8c231d7ac0714695dd45ae68cbee506d602700548af",
213
+ "name": "normals_matterport3d",
214
+ "path": "<local-path>",
215
+ "sample_list_sha256": "37280ca662f240992afc5fc0400f1a5d27d4bbb08f61665bd46ac3d27cb0e5ff",
216
+ "sampling": "scene_balanced",
217
+ "source_dataset": "matterport3d",
218
+ "task": "surface_normal_estimation",
219
+ "type": "indexed_prepared_tar",
220
+ "validation_report": "<local-path>",
221
+ "weight": 0.04
222
+ },
223
+ {
224
+ "archive_sha256": "bb0c1653e5802890d2cc7358941750d9020d38081b9d00882d84ec832efed501",
225
+ "geometry_version": "geometry_canonical_v2",
226
+ "index_path": "<local-path>",
227
+ "index_sha256": "082e88258e104d1e8c96e1abaef8991506cd2ce4951b1c402ad6fab439ed5feb",
228
+ "name": "normals_replica",
229
+ "path": "<local-path>",
230
+ "sample_list_sha256": "89bcff6ca51ff82624ff44aa986bf6982fe7863195f63d4865e889c8560700bb",
231
+ "sampling": "scene_balanced",
232
+ "source_dataset": "replica",
233
+ "task": "surface_normal_estimation",
234
+ "type": "indexed_prepared_tar",
235
+ "validation_report": "<local-path>",
236
+ "weight": 0.04
237
+ },
238
+ {
239
+ "archive_sha256": "d4e70dbeb5ef281be17a304ab89f7253cec8d2ba716d307e855c788d57b125e3",
240
+ "geometry_version": "geometry_canonical_v2",
241
+ "index_path": "<local-path>",
242
+ "index_sha256": "10d2f4bbf5e698f60657f9e0e06eb6d1d6319939d2569d89983503981dfbb93e",
243
+ "name": "normals_synscapes",
244
+ "path": "<local-path>",
245
+ "sample_list_sha256": "7118294ad359f2cb9c1acb6a93719a3a6769f3dbaa4b646a216f8788de961520",
246
+ "sampling": "row",
247
+ "source_dataset": "synscapes",
248
+ "task": "surface_normal_estimation",
249
+ "type": "indexed_prepared_tar",
250
+ "validation_report": "<local-path>",
251
+ "weight": 0.04
252
+ },
253
+ {
254
+ "archive_sha256": "a3550f40fcfe60a72622cd3f1cef38f5eb4870c9653a50a740a3c679d31accb4",
255
+ "geometry_version": "geometry_vision_banana_v3",
256
+ "index_path": "<local-path>",
257
+ "index_sha256": "0f01831b926d04216e173624146a693e96d95728a9bd234a974313400d97bae5",
258
+ "name": "normals_taskonomy",
259
+ "path": "<local-path>",
260
+ "sample_list_sha256": "698aad7a87803c74ff750bfaa2cd37fbdf29d4b5b73e77009aea4b41c424c7d2",
261
+ "sampling": "scene_balanced",
262
+ "source_dataset": "taskonomy",
263
+ "task": "surface_normal_estimation",
264
+ "type": "indexed_prepared_tar",
265
+ "validation_report": "<local-path>",
266
+ "weight": 0.04
267
+ }
268
+ ],
269
+ "excluded_geometry_sources": [],
270
+ "experiment_id": "e8_ratio_depth10_normals20",
271
+ "profile": "geometry_ablation_20260801_v1",
272
+ "samples_per_epoch": 160000,
273
+ "schema_version": 2,
274
+ "seed": 42,
275
+ "shuffle": true,
276
+ "source_weighting": "equal_within_geometry_task_after_exclusion",
277
+ "task_group_weights": {
278
+ "depth": 0.1,
279
+ "non_geometry": 0.7,
280
+ "surface_normals": 0.2
281
+ }
282
+ }
ablation_models/additional_studies/geometry/depth_normal_allocation/depth10_normals20/checkpoint-37500/dataset_specs.resolved.json ADDED
@@ -0,0 +1,202 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "datasets": [
3
+ {
4
+ "filters": {
5
+ "source_dataset": "acdc"
6
+ },
7
+ "index_path": "<local-path>",
8
+ "name": "semseg_acdc",
9
+ "path": "<local-path>",
10
+ "sampling": "row",
11
+ "type": "indexed_prepared_tar",
12
+ "weight": 0.035
13
+ },
14
+ {
15
+ "filters": {
16
+ "source_dataset": "bdd"
17
+ },
18
+ "index_path": "<local-path>",
19
+ "name": "semseg_bdd",
20
+ "path": "<local-path>",
21
+ "sampling": "row",
22
+ "type": "indexed_prepared_tar",
23
+ "weight": 0.035
24
+ },
25
+ {
26
+ "filters": {
27
+ "source_dataset": "ade20k"
28
+ },
29
+ "index_path": "<local-path>",
30
+ "name": "semseg_ade20k",
31
+ "path": "<local-path>",
32
+ "sampling": "row",
33
+ "type": "indexed_prepared_tar",
34
+ "weight": 0.035
35
+ },
36
+ {
37
+ "filters": {
38
+ "source_dataset": "voc"
39
+ },
40
+ "index_path": "<local-path>",
41
+ "name": "semseg_voc",
42
+ "path": "<local-path>",
43
+ "sampling": "row",
44
+ "type": "indexed_prepared_tar",
45
+ "weight": 0.035
46
+ },
47
+ {
48
+ "filters": {
49
+ "source_dataset": "ade20k_inst"
50
+ },
51
+ "index_path": "<local-path>",
52
+ "name": "instanceseg_ade20k",
53
+ "path": "<local-path>",
54
+ "sampling": "row",
55
+ "type": "indexed_prepared_tar",
56
+ "weight": 0.032
57
+ },
58
+ {
59
+ "filters": {
60
+ "source_dataset": "mapillary"
61
+ },
62
+ "index_path": "<local-path>",
63
+ "name": "instanceseg_mapillary",
64
+ "path": "<local-path>",
65
+ "sampling": "row",
66
+ "type": "indexed_prepared_tar",
67
+ "weight": 0.032
68
+ },
69
+ {
70
+ "filters": {
71
+ "source_dataset": "voc_inst"
72
+ },
73
+ "index_path": "<local-path>",
74
+ "name": "instanceseg_voc",
75
+ "path": "<local-path>",
76
+ "sampling": "row",
77
+ "type": "indexed_prepared_tar",
78
+ "weight": 0.016
79
+ },
80
+ {
81
+ "index_path": "<local-path>",
82
+ "name": "refseg",
83
+ "path": "<local-path>",
84
+ "sampling": "row",
85
+ "type": "indexed_prepared_tar",
86
+ "weight": 0.08
87
+ },
88
+ {
89
+ "index_path": "<local-path>",
90
+ "name": "pico_banana",
91
+ "path": "<local-path>",
92
+ "sampling": "row",
93
+ "type": "indexed_prepared_tar",
94
+ "weight": 0.15
95
+ },
96
+ {
97
+ "index_path": "<local-path>",
98
+ "name": "identity",
99
+ "path": "<local-path>",
100
+ "sampling": "row",
101
+ "type": "indexed_prepared_tar",
102
+ "weight": 0.1
103
+ },
104
+ {
105
+ "name": "pd12m",
106
+ "shards": "<local-path>",
107
+ "source_dataset": "pd12m",
108
+ "streaming": {
109
+ "error_policy": "warn_and_continue",
110
+ "resampled": true,
111
+ "sample_shuffle_buffer": 10000,
112
+ "shard_shuffle": true
113
+ },
114
+ "task": "text_to_image",
115
+ "type": "webdataset_text_to_image",
116
+ "weight": 0.15
117
+ },
118
+ {
119
+ "index_path": "<local-path>",
120
+ "name": "depth_hypersim",
121
+ "path": "<local-path>",
122
+ "sampling": "scene_balanced",
123
+ "type": "indexed_prepared_tar",
124
+ "weight": 0.02
125
+ },
126
+ {
127
+ "index_path": "<local-path>",
128
+ "name": "depth_matterport3d",
129
+ "path": "<local-path>",
130
+ "sampling": "scene_balanced",
131
+ "type": "indexed_prepared_tar",
132
+ "weight": 0.02
133
+ },
134
+ {
135
+ "index_path": "<local-path>",
136
+ "name": "depth_replica",
137
+ "path": "<local-path>",
138
+ "sampling": "scene_balanced",
139
+ "type": "indexed_prepared_tar",
140
+ "weight": 0.02
141
+ },
142
+ {
143
+ "index_path": "<local-path>",
144
+ "name": "depth_synscapes",
145
+ "path": "<local-path>",
146
+ "sampling": "row",
147
+ "type": "indexed_prepared_tar",
148
+ "weight": 0.02
149
+ },
150
+ {
151
+ "index_path": "<local-path>",
152
+ "name": "depth_taskonomy",
153
+ "path": "<local-path>",
154
+ "sampling": "scene_balanced",
155
+ "type": "indexed_prepared_tar",
156
+ "weight": 0.02
157
+ },
158
+ {
159
+ "index_path": "<local-path>",
160
+ "name": "normals_hypersim",
161
+ "path": "<local-path>",
162
+ "sampling": "scene_balanced",
163
+ "type": "indexed_prepared_tar",
164
+ "weight": 0.04
165
+ },
166
+ {
167
+ "index_path": "<local-path>",
168
+ "name": "normals_matterport3d",
169
+ "path": "<local-path>",
170
+ "sampling": "scene_balanced",
171
+ "type": "indexed_prepared_tar",
172
+ "weight": 0.04
173
+ },
174
+ {
175
+ "index_path": "<local-path>",
176
+ "name": "normals_replica",
177
+ "path": "<local-path>",
178
+ "sampling": "scene_balanced",
179
+ "type": "indexed_prepared_tar",
180
+ "weight": 0.04
181
+ },
182
+ {
183
+ "index_path": "<local-path>",
184
+ "name": "normals_synscapes",
185
+ "path": "<local-path>",
186
+ "sampling": "row",
187
+ "type": "indexed_prepared_tar",
188
+ "weight": 0.04
189
+ },
190
+ {
191
+ "index_path": "<local-path>",
192
+ "name": "normals_taskonomy",
193
+ "path": "<local-path>",
194
+ "sampling": "scene_balanced",
195
+ "type": "indexed_prepared_tar",
196
+ "weight": 0.04
197
+ }
198
+ ],
199
+ "samples_per_epoch": 160000,
200
+ "seed": 42,
201
+ "shuffle": true
202
+ }
ablation_models/additional_studies/geometry/depth_normal_allocation/depth10_normals20/checkpoint-37500/evaluation_protocol.json ADDED
@@ -0,0 +1,7 @@
 
 
 
 
 
 
 
 
1
+ {
2
+ "scope": "Trained ablation checkpoint; consult the cited study for datasets and inference settings.",
3
+ "corrected_depth": "Marigold-v2 PPD per-image log-depth RANSAC alignment",
4
+ "corrected_normals": "geometry-v6 assets and fixed camera-facing decoder; no test-set axis fitting",
5
+ "historical_results": "Historical tables use different assets, datasets or scoring conventions; do not mix with corrected scores.",
6
+ "checkpoint_selection": "Evaluated checkpoint; intermediate training-duration and warmup checkpoints are retained."
7
+ }
ablation_models/additional_studies/geometry/depth_normal_allocation/depth10_normals20/checkpoint-37500/run_spec.json ADDED
@@ -0,0 +1,91 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "account": "h200_y506dd10",
3
+ "campaign_id": "geometry_klein_37500_continuations_20260811_v1",
4
+ "continuation": {
5
+ "mode": "exact_optimizer_scheduler_rng_sampler_resume",
6
+ "source_checkpoint": "<local-path>",
7
+ "source_checkpoint_step": 20000,
8
+ "source_dataset_specs_sha256": "548b407be93526bd1aa1bd29385668db39e577bd9e217eb351d5bd37a3e842ca",
9
+ "source_run": "<local-path>",
10
+ "source_run_spec_sha256": "95de623c86ba93259354882ac0ca5c230f3311ae76881d6d75fa7bebd1100df5",
11
+ "target_checkpoint_step": 37500
12
+ },
13
+ "environment": {
14
+ "ALLOW_LEGACY_HYPERSIM_RANGE_DEPTH": "0",
15
+ "ALLOW_UNVERIFIED_NORMAL_SOURCES": "0",
16
+ "ASPECT_RATIO_BUCKETS": "512,512;576,448;448,576;608,416;416,608",
17
+ "AUTO_RECONCILE_ON_COMPLETE": "0",
18
+ "CACHE_LATENTS": "0",
19
+ "CHECKPOINTING_STEPS": "500",
20
+ "CHECKPOINTS_TOTAL_LIMIT": "3",
21
+ "CUDA_VISIBLE_DEVICES": "0,1,2,3",
22
+ "DATASET_SPECS": "<local-path>",
23
+ "DEPTH_ENCODING": "relative_inverse_grayscale",
24
+ "DEPTH_PROMPT_CONTRACT": "campaign_detailed_inverse",
25
+ "DEPTH_TARGET_RESIZE_POLICY": "metric_field_then_encode",
26
+ "DISTRIBUTED_STRATEGY": "fsdp",
27
+ "FSDP_TRANSFORMER_DTYPE_POLICY": "fp32_all",
28
+ "GEOMETRY_TASKS": "depth,normals",
29
+ "GRADIENT_ACCUMULATION_STEPS": "2",
30
+ "GUIDANCE_SCALE": "3.5",
31
+ "LEARNING_RATE": "2e-06",
32
+ "LOSS_MASK_EROSION_RADIUS": "0",
33
+ "LOSS_MASK_MIN_VALID_FRACTION": "0.5",
34
+ "LOSS_MASK_MODE": "hard",
35
+ "LOSS_MASK_POLICY": "auto",
36
+ "LOSS_MASK_REDUCTION": "per_sample",
37
+ "LR_SCHEDULER": "constant_with_warmup",
38
+ "LR_WARMUP_STEPS": "1000",
39
+ "MAX_TRAIN_STEPS": "37500",
40
+ "MIXED_PRECISION": "bf16",
41
+ "MODEL": "<local-path>",
42
+ "MODEL_FAMILY": "flux2_klein",
43
+ "NORMAL_PROMPT_CONTRACT": "campaign_short",
44
+ "NORMAL_TARGET_ENCODING": "vision_banana_x_left_red",
45
+ "NORMAL_TARGET_RESIZE_POLICY": "vector_field_renormalize_then_encode",
46
+ "NUM_PROCESSES": "4",
47
+ "OUTPUT_DIR": "<local-path>",
48
+ "PERMANENT_CHECKPOINT_STEPS": "20000,37500",
49
+ "PROMPT_ENCODING_MODE": "auto",
50
+ "REPORT_TO": "tensorboard",
51
+ "RESAMPLE_DATASET_EACH_EPOCH": "1",
52
+ "RESOLUTION": "512",
53
+ "RESUME_FROM_CHECKPOINT": "latest",
54
+ "ROOT": "<local-path>",
55
+ "RUN_DIR": "<local-path>",
56
+ "RUN_ID": "c375_ratio_depth10_normals20",
57
+ "SAMPLES_PER_EPOCH": "160000",
58
+ "SEED": "42",
59
+ "SKIP_FINAL_INFERENCE": "1",
60
+ "SKIP_FINAL_SAVE": "1",
61
+ "TASK_HOMOGENEOUS_BATCHES": "1",
62
+ "TRAINING_SUITE": "geometry_ablation_20260801_v1",
63
+ "TRAIN_ALL_TRANSFORMER_BLOCKS": "1",
64
+ "TRAIN_BATCH_SIZE": "1",
65
+ "TRAIN_FULL_TRANSFORMER": "0",
66
+ "TRAIN_INPUT_LAYERS": "0",
67
+ "TRAIN_LAST_SINGLE_TRANSFORMER_BLOCKS": "0",
68
+ "TRAIN_LAST_TRANSFORMER_BLOCKS": "0",
69
+ "TRAIN_METHOD": "partial_full_rank",
70
+ "TRAIN_NON_BLOCK_LAYERS": "1",
71
+ "TRAIN_OUTPUT_LAYERS": "1",
72
+ "WEIGHT_DECAY": "1e-4"
73
+ },
74
+ "evaluate_checkpoints": [
75
+ 37500
76
+ ],
77
+ "evaluation_scope": "geometry_and_compact_guardrails",
78
+ "evaluation_suites": [
79
+ "geometry",
80
+ "semantic",
81
+ "refseg",
82
+ "instance",
83
+ "imgedit"
84
+ ],
85
+ "group": "downstream",
86
+ "run_id": "c375_ratio_depth10_normals20",
87
+ "schema_version": 1,
88
+ "source_campaign_id": "geometry_ablation_20260801_v1",
89
+ "source_run_id": "e8_ratio_depth10_normals20",
90
+ "status": "PLANNED"
91
+ }
ablation_models/additional_studies/geometry/depth_normal_allocation/depth10_normals20/checkpoint-37500/training_args.json ADDED
@@ -0,0 +1,129 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "adam_beta1": 0.9,
3
+ "adam_beta2": 0.999,
4
+ "adam_epsilon": 1e-08,
5
+ "adam_weight_decay": 0.0001,
6
+ "adam_weight_decay_text_encoder": 0.001,
7
+ "allow_large_prompt_precompute_cache": false,
8
+ "allow_legacy_hypersim_range_depth": false,
9
+ "allow_tf32": false,
10
+ "allow_unverified_normal_sources": false,
11
+ "angular_normal_aux_weight": 0.0,
12
+ "aspect_ratio_buckets": "512,512;576,448;448,576;608,416;416,608",
13
+ "bnb_quantization_config_path": null,
14
+ "cache_dir": null,
15
+ "cache_latents": false,
16
+ "caption_column": "text",
17
+ "center_crop": false,
18
+ "checkpointing_steps": 500,
19
+ "checkpoints_total_limit": 3,
20
+ "class_data_dir": null,
21
+ "cond_image_column": "conditioning_image",
22
+ "dataloader_num_workers": 0,
23
+ "dataset_config_name": null,
24
+ "dataset_index_num_workers": 16,
25
+ "dataset_name": null,
26
+ "dataset_pool_seed": 42,
27
+ "dataset_specs": "<local-path>",
28
+ "depth_prompt_contract": "campaign_detailed_inverse",
29
+ "depth_target_encoding": "relative_inverse_grayscale",
30
+ "depth_target_resize_policy": "metric_field_then_encode",
31
+ "do_fp8_training": false,
32
+ "enable_npu_flash_attention": false,
33
+ "final_validation_prompt": null,
34
+ "fsdp_text_encoder": false,
35
+ "fsdp_transformer": true,
36
+ "fsdp_transformer_cls_to_wrap": "Flux2TransformerBlock,Flux2SingleTransformerBlock",
37
+ "fsdp_transformer_dtype_policy": "fp32_all",
38
+ "gradient_accumulation_steps": 2,
39
+ "gradient_checkpointing": true,
40
+ "guidance_scale": 3.5,
41
+ "hub_model_id": null,
42
+ "hub_token": null,
43
+ "image_column": "image",
44
+ "instance_data_dir": null,
45
+ "instance_prompt": "[TASK: IMAGE_EDITING] Edit the image according to the instruction.",
46
+ "learning_rate": 2e-06,
47
+ "local_rank": 0,
48
+ "logging_dir": "logs",
49
+ "logit_mean": 0.0,
50
+ "logit_std": 1.0,
51
+ "lora_alpha": 64,
52
+ "lora_dropout": 0.0,
53
+ "lora_layers": null,
54
+ "lora_rank": 64,
55
+ "loss_mask_erosion_radius": 0,
56
+ "loss_mask_min_sample_valid_fraction": 0.0,
57
+ "loss_mask_min_valid_fraction": 0.5,
58
+ "loss_mask_mode": "hard",
59
+ "loss_mask_policy": "auto",
60
+ "loss_mask_reduction": "per_sample",
61
+ "lr_num_cycles": 1,
62
+ "lr_power": 1.0,
63
+ "lr_scheduler": "constant_with_warmup",
64
+ "lr_warmup_steps": 1000,
65
+ "max_grad_norm": 1.0,
66
+ "max_sequence_length": 512,
67
+ "max_train_steps": 37500,
68
+ "metric_depth_aux_max_depth": 80.0,
69
+ "metric_depth_aux_weight": 0.0,
70
+ "mixed_precision": "bf16",
71
+ "mode_scale": 1.29,
72
+ "normal_prompt_contract": "campaign_short",
73
+ "normal_target_encoding": "vision_banana_x_left_red",
74
+ "normal_target_resize_policy": "vector_field_renormalize_then_encode",
75
+ "num_train_epochs": 1,
76
+ "num_validation_images": 4,
77
+ "offload": false,
78
+ "optimizer": "AdamW",
79
+ "output_dir": "<local-path>",
80
+ "palette_manifold_loss_weight": 0.0,
81
+ "permanent_checkpoint_steps": "20000,37500",
82
+ "precompute_cache_dir": null,
83
+ "pretrained_model_name_or_path": "<local-path>",
84
+ "prodigy_beta3": null,
85
+ "prodigy_decouple": true,
86
+ "prodigy_safeguard_warmup": true,
87
+ "prodigy_use_bias_correction": true,
88
+ "prompt_embedding_batch_size": 16,
89
+ "prompt_encoding_mode": "auto",
90
+ "prompt_precompute_max_unique": 2048,
91
+ "push_to_hub": false,
92
+ "random_flip": false,
93
+ "repeats": 1,
94
+ "report_to": "tensorboard",
95
+ "resample_dataset_each_epoch": true,
96
+ "resolution": 512,
97
+ "resume_from_checkpoint": "latest",
98
+ "revision": null,
99
+ "sample_batch_size": 4,
100
+ "samples_per_epoch": 160000,
101
+ "scale_lr": false,
102
+ "seed": 42,
103
+ "skip_final_inference": true,
104
+ "skip_final_save": true,
105
+ "task_homogeneous_batches": true,
106
+ "train_all_transformer_blocks": true,
107
+ "train_batch_size": 1,
108
+ "train_full_transformer": false,
109
+ "train_input_layers": false,
110
+ "train_last_single_transformer_blocks": 0,
111
+ "train_last_transformer_blocks": 0,
112
+ "train_method": "partial_full_rank",
113
+ "train_non_block_layers": true,
114
+ "train_output_layers": true,
115
+ "train_single_transformer_block_indices": null,
116
+ "train_transformer_block_indices": null,
117
+ "unit_normal_loss_weight": 0.0,
118
+ "upcast_before_saving": false,
119
+ "use_8bit_adam": false,
120
+ "validation_epochs": 50,
121
+ "validation_height": null,
122
+ "validation_image": null,
123
+ "validation_num_inference_steps": 28,
124
+ "validation_prompt": null,
125
+ "validation_prompts_file": null,
126
+ "validation_width": null,
127
+ "variant": null,
128
+ "weighting_scheme": "none"
129
+ }