Image-to-Image
Diffusers
Safetensors
flux
semantic-segmentation
instance-segmentation
referring-segmentation
depth-estimation
surface-normal-estimation
Instructions to use FunAILab/Open-Vision-Banana with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use FunAILab/Open-Vision-Banana with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline from diffusers.utils import load_image # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("FunAILab/Open-Vision-Banana", dtype=torch.bfloat16, device_map="cuda") prompt = "Turn this cat into a dog" input_image = load_image("https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/cat.png") image = pipe(image=input_image, prompt=prompt).images[0] - Notebooks
- Google Colab
- Kaggle
Add training metadata and checksums: creative_fraction/generation0
Browse files- ablation_models/additional_studies/cross_task/creative_task_fraction/generation0/checkpoint-50000/COMPLETE.json +7 -0
- ablation_models/additional_studies/cross_task/creative_task_fraction/generation0/checkpoint-50000/SHA256SUMS +6 -0
- ablation_models/additional_studies/cross_task/creative_task_fraction/generation0/checkpoint-50000/checksum_validation.json +30 -0
- ablation_models/additional_studies/cross_task/creative_task_fraction/generation0/checkpoint-50000/dataset_specs.json +213 -0
- ablation_models/additional_studies/cross_task/creative_task_fraction/generation0/checkpoint-50000/dataset_specs.resolved.json +261 -0
- ablation_models/additional_studies/cross_task/creative_task_fraction/generation0/checkpoint-50000/evaluation_protocol.json +7 -0
- ablation_models/additional_studies/cross_task/creative_task_fraction/generation0/checkpoint-50000/run_spec.json +112 -0
- ablation_models/additional_studies/cross_task/creative_task_fraction/generation0/checkpoint-50000/training_args.json +129 -0
ablation_models/additional_studies/cross_task/creative_task_fraction/generation0/checkpoint-50000/COMPLETE.json
ADDED
|
@@ -0,0 +1,7 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"status": "COMPLETE",
|
| 3 |
+
"attempt": "3",
|
| 4 |
+
"completed_utc": "2026-08-26T00:29:38.720692+00:00",
|
| 5 |
+
"completion_source": "trainer_exit_zero",
|
| 6 |
+
"wrapper_exit_code": 0
|
| 7 |
+
}
|
ablation_models/additional_studies/cross_task/creative_task_fraction/generation0/checkpoint-50000/SHA256SUMS
ADDED
|
@@ -0,0 +1,6 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
97ea39929ca88b293ce307114cfc265a8345d82675b69b3232269390cfdf09ee config.json
|
| 2 |
+
235a30e4d3ac93b6eaf483fe690af3d302d18b9e883878695fa60aff5ff0818b diffusion_pytorch_model.safetensors.index.json
|
| 3 |
+
75ceb6bcc6f1fb681f78775d6b2f008c6df8b69056aea8c62e9ca7705c7004d3 diffusion_pytorch_model-00001-of-00004.safetensors
|
| 4 |
+
27872f67b53cf95744c998d6fb1f6b68e11c68e67f25b74392c0867b570bbb07 diffusion_pytorch_model-00002-of-00004.safetensors
|
| 5 |
+
8ff713d7c81a5d9f6c042f94018ba125f73ac5e27cb0900d96ff72c9530d2bc8 diffusion_pytorch_model-00003-of-00004.safetensors
|
| 6 |
+
f49b8228ddccbe9c8909084dd75619964660c29b25bf51b4da9f42925e7d6c85 diffusion_pytorch_model-00004-of-00004.safetensors
|
ablation_models/additional_studies/cross_task/creative_task_fraction/generation0/checkpoint-50000/checksum_validation.json
ADDED
|
@@ -0,0 +1,30 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"fingerprints": {
|
| 3 |
+
"config.json": {
|
| 4 |
+
"size": 552,
|
| 5 |
+
"mtime_ns": 1787704144000000000
|
| 6 |
+
},
|
| 7 |
+
"diffusion_pytorch_model.safetensors.index.json": {
|
| 8 |
+
"size": 24875,
|
| 9 |
+
"mtime_ns": 1787704173000000000
|
| 10 |
+
},
|
| 11 |
+
"diffusion_pytorch_model-00001-of-00004.safetensors": {
|
| 12 |
+
"size": 9804206048,
|
| 13 |
+
"mtime_ns": 1787704152000000000
|
| 14 |
+
},
|
| 15 |
+
"diffusion_pytorch_model-00002-of-00004.safetensors": {
|
| 16 |
+
"size": 9797913704,
|
| 17 |
+
"mtime_ns": 1787704159000000000
|
| 18 |
+
},
|
| 19 |
+
"diffusion_pytorch_model-00003-of-00004.safetensors": {
|
| 20 |
+
"size": 9596584288,
|
| 21 |
+
"mtime_ns": 1787704167000000000
|
| 22 |
+
},
|
| 23 |
+
"diffusion_pytorch_model-00004-of-00004.safetensors": {
|
| 24 |
+
"size": 7115649136,
|
| 25 |
+
"mtime_ns": 1787704173000000000
|
| 26 |
+
}
|
| 27 |
+
},
|
| 28 |
+
"sha256_manifest": "baabfc5559ee0269be67c13768972c506c501444a655b8a658404887b1b268fb",
|
| 29 |
+
"validation": "PASS"
|
| 30 |
+
}
|
ablation_models/additional_studies/cross_task/creative_task_fraction/generation0/checkpoint-50000/dataset_specs.json
ADDED
|
@@ -0,0 +1,213 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"campaign_id": "geometry_phase2_20260806_v1",
|
| 3 |
+
"datasets": [
|
| 4 |
+
{
|
| 5 |
+
"filters": {
|
| 6 |
+
"source_dataset": "acdc"
|
| 7 |
+
},
|
| 8 |
+
"index_path": "<local-path>",
|
| 9 |
+
"name": "semseg_acdc",
|
| 10 |
+
"path": "<local-path>",
|
| 11 |
+
"sampling": "row",
|
| 12 |
+
"type": "indexed_prepared_tar",
|
| 13 |
+
"weight": 0.04125
|
| 14 |
+
},
|
| 15 |
+
{
|
| 16 |
+
"filters": {
|
| 17 |
+
"source_dataset": "bdd"
|
| 18 |
+
},
|
| 19 |
+
"index_path": "<local-path>",
|
| 20 |
+
"name": "semseg_bdd",
|
| 21 |
+
"path": "<local-path>",
|
| 22 |
+
"sampling": "row",
|
| 23 |
+
"type": "indexed_prepared_tar",
|
| 24 |
+
"weight": 0.04125
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"filters": {
|
| 28 |
+
"source_dataset": "ade20k"
|
| 29 |
+
},
|
| 30 |
+
"index_path": "<local-path>",
|
| 31 |
+
"name": "semseg_ade20k",
|
| 32 |
+
"path": "<local-path>",
|
| 33 |
+
"sampling": "row",
|
| 34 |
+
"type": "indexed_prepared_tar",
|
| 35 |
+
"weight": 0.04125
|
| 36 |
+
},
|
| 37 |
+
{
|
| 38 |
+
"filters": {
|
| 39 |
+
"source_dataset": "voc"
|
| 40 |
+
},
|
| 41 |
+
"index_path": "<local-path>",
|
| 42 |
+
"name": "semseg_voc",
|
| 43 |
+
"path": "<local-path>",
|
| 44 |
+
"sampling": "row",
|
| 45 |
+
"type": "indexed_prepared_tar",
|
| 46 |
+
"weight": 0.04125
|
| 47 |
+
},
|
| 48 |
+
{
|
| 49 |
+
"filters": {
|
| 50 |
+
"source_dataset": "ade20k_inst"
|
| 51 |
+
},
|
| 52 |
+
"index_path": "<local-path>",
|
| 53 |
+
"name": "instanceseg_ade20k",
|
| 54 |
+
"path": "<local-path>",
|
| 55 |
+
"sampling": "row",
|
| 56 |
+
"type": "indexed_prepared_tar",
|
| 57 |
+
"weight": 0.042
|
| 58 |
+
},
|
| 59 |
+
{
|
| 60 |
+
"filters": {
|
| 61 |
+
"source_dataset": "mapillary"
|
| 62 |
+
},
|
| 63 |
+
"index_path": "<local-path>",
|
| 64 |
+
"name": "instanceseg_mapillary",
|
| 65 |
+
"path": "<local-path>",
|
| 66 |
+
"sampling": "row",
|
| 67 |
+
"type": "indexed_prepared_tar",
|
| 68 |
+
"weight": 0.042
|
| 69 |
+
},
|
| 70 |
+
{
|
| 71 |
+
"filters": {
|
| 72 |
+
"source_dataset": "voc_inst"
|
| 73 |
+
},
|
| 74 |
+
"index_path": "<local-path>",
|
| 75 |
+
"name": "instanceseg_voc",
|
| 76 |
+
"path": "<local-path>",
|
| 77 |
+
"sampling": "row",
|
| 78 |
+
"type": "indexed_prepared_tar",
|
| 79 |
+
"weight": 0.021
|
| 80 |
+
},
|
| 81 |
+
{
|
| 82 |
+
"index_path": "<local-path>",
|
| 83 |
+
"name": "refseg",
|
| 84 |
+
"path": "<local-path>",
|
| 85 |
+
"sampling": "row",
|
| 86 |
+
"type": "indexed_prepared_tar",
|
| 87 |
+
"weight": 0.105
|
| 88 |
+
},
|
| 89 |
+
{
|
| 90 |
+
"index_path": "<local-path>",
|
| 91 |
+
"name": "pico_banana",
|
| 92 |
+
"path": "<local-path>",
|
| 93 |
+
"sampling": "row",
|
| 94 |
+
"type": "indexed_prepared_tar",
|
| 95 |
+
"weight": 0.165
|
| 96 |
+
},
|
| 97 |
+
{
|
| 98 |
+
"index_path": "<local-path>",
|
| 99 |
+
"name": "identity",
|
| 100 |
+
"path": "<local-path>",
|
| 101 |
+
"sampling": "row",
|
| 102 |
+
"type": "indexed_prepared_tar",
|
| 103 |
+
"weight": 0.11
|
| 104 |
+
},
|
| 105 |
+
{
|
| 106 |
+
"index_path": "<local-path>",
|
| 107 |
+
"name": "depth_hypersim",
|
| 108 |
+
"path": "<local-path>",
|
| 109 |
+
"sampling": "scene_balanced",
|
| 110 |
+
"type": "indexed_prepared_tar",
|
| 111 |
+
"weight": 0.045
|
| 112 |
+
},
|
| 113 |
+
{
|
| 114 |
+
"index_path": "<local-path>",
|
| 115 |
+
"name": "depth_matterport3d",
|
| 116 |
+
"path": "<local-path>",
|
| 117 |
+
"sampling": "scene_balanced",
|
| 118 |
+
"type": "indexed_prepared_tar",
|
| 119 |
+
"weight": 0.045
|
| 120 |
+
},
|
| 121 |
+
{
|
| 122 |
+
"index_path": "<local-path>",
|
| 123 |
+
"name": "depth_replica",
|
| 124 |
+
"path": "<local-path>",
|
| 125 |
+
"sampling": "scene_balanced",
|
| 126 |
+
"type": "indexed_prepared_tar",
|
| 127 |
+
"weight": 0.045
|
| 128 |
+
},
|
| 129 |
+
{
|
| 130 |
+
"index_path": "<local-path>",
|
| 131 |
+
"name": "depth_synscapes",
|
| 132 |
+
"path": "<local-path>",
|
| 133 |
+
"sampling": "row",
|
| 134 |
+
"type": "indexed_prepared_tar",
|
| 135 |
+
"weight": 0.045
|
| 136 |
+
},
|
| 137 |
+
{
|
| 138 |
+
"index_path": "<local-path>",
|
| 139 |
+
"name": "depth_taskonomy",
|
| 140 |
+
"path": "<local-path>",
|
| 141 |
+
"sampling": "scene_balanced",
|
| 142 |
+
"type": "indexed_prepared_tar",
|
| 143 |
+
"weight": 0.045
|
| 144 |
+
},
|
| 145 |
+
{
|
| 146 |
+
"index_path": "<local-path>",
|
| 147 |
+
"name": "normals_hypersim",
|
| 148 |
+
"path": "<local-path>",
|
| 149 |
+
"sampling": "scene_balanced",
|
| 150 |
+
"type": "indexed_prepared_tar",
|
| 151 |
+
"weight": 0.025
|
| 152 |
+
},
|
| 153 |
+
{
|
| 154 |
+
"index_path": "<local-path>",
|
| 155 |
+
"name": "normals_matterport3d",
|
| 156 |
+
"path": "<local-path>",
|
| 157 |
+
"sampling": "scene_balanced",
|
| 158 |
+
"type": "indexed_prepared_tar",
|
| 159 |
+
"weight": 0.025
|
| 160 |
+
},
|
| 161 |
+
{
|
| 162 |
+
"index_path": "<local-path>",
|
| 163 |
+
"name": "normals_replica",
|
| 164 |
+
"path": "<local-path>",
|
| 165 |
+
"sampling": "scene_balanced",
|
| 166 |
+
"type": "indexed_prepared_tar",
|
| 167 |
+
"weight": 0.025
|
| 168 |
+
},
|
| 169 |
+
{
|
| 170 |
+
"index_path": "<local-path>",
|
| 171 |
+
"name": "normals_synscapes",
|
| 172 |
+
"path": "<local-path>",
|
| 173 |
+
"sampling": "row",
|
| 174 |
+
"type": "indexed_prepared_tar",
|
| 175 |
+
"weight": 0.025
|
| 176 |
+
},
|
| 177 |
+
{
|
| 178 |
+
"index_path": "<local-path>",
|
| 179 |
+
"name": "normals_taskonomy",
|
| 180 |
+
"path": "<local-path>",
|
| 181 |
+
"sampling": "scene_balanced",
|
| 182 |
+
"type": "indexed_prepared_tar",
|
| 183 |
+
"weight": 0.025
|
| 184 |
+
}
|
| 185 |
+
],
|
| 186 |
+
"excluded_geometry_sources": [],
|
| 187 |
+
"geometry_version": "geometry_canonical_v2_v3",
|
| 188 |
+
"planned_task_group_samples": {
|
| 189 |
+
"depth": 36000,
|
| 190 |
+
"identity": 17600,
|
| 191 |
+
"instance": 16800,
|
| 192 |
+
"pd12m": 0,
|
| 193 |
+
"pico": 26400,
|
| 194 |
+
"refseg": 16800,
|
| 195 |
+
"semantic": 26400,
|
| 196 |
+
"surface_normals": 20000
|
| 197 |
+
},
|
| 198 |
+
"samples_per_epoch": 160000,
|
| 199 |
+
"schema_version": 2,
|
| 200 |
+
"seed": 42,
|
| 201 |
+
"shuffle": true,
|
| 202 |
+
"source_weighting": "declared_relative_weight_within_task",
|
| 203 |
+
"task_group_weights": {
|
| 204 |
+
"depth": 0.225,
|
| 205 |
+
"identity": 0.11,
|
| 206 |
+
"instance": 0.105,
|
| 207 |
+
"pd12m": 0.0,
|
| 208 |
+
"pico": 0.165,
|
| 209 |
+
"refseg": 0.105,
|
| 210 |
+
"semantic": 0.165,
|
| 211 |
+
"surface_normals": 0.125
|
| 212 |
+
}
|
| 213 |
+
}
|
ablation_models/additional_studies/cross_task/creative_task_fraction/generation0/checkpoint-50000/dataset_specs.resolved.json
ADDED
|
@@ -0,0 +1,261 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"datasets": [
|
| 3 |
+
{
|
| 4 |
+
"filters": {
|
| 5 |
+
"source_dataset": "acdc"
|
| 6 |
+
},
|
| 7 |
+
"index_path": "<local-path>",
|
| 8 |
+
"name": "semseg_acdc",
|
| 9 |
+
"num_rows": 1600,
|
| 10 |
+
"path": "<local-path>",
|
| 11 |
+
"sampling": "row",
|
| 12 |
+
"type": "indexed_prepared_tar",
|
| 13 |
+
"unique_data_fraction": 1.0,
|
| 14 |
+
"unique_data_seed": 42,
|
| 15 |
+
"weight": 0.04125
|
| 16 |
+
},
|
| 17 |
+
{
|
| 18 |
+
"filters": {
|
| 19 |
+
"source_dataset": "bdd"
|
| 20 |
+
},
|
| 21 |
+
"index_path": "<local-path>",
|
| 22 |
+
"name": "semseg_bdd",
|
| 23 |
+
"num_rows": 7000,
|
| 24 |
+
"path": "<local-path>",
|
| 25 |
+
"sampling": "row",
|
| 26 |
+
"type": "indexed_prepared_tar",
|
| 27 |
+
"unique_data_fraction": 1.0,
|
| 28 |
+
"unique_data_seed": 42,
|
| 29 |
+
"weight": 0.04125
|
| 30 |
+
},
|
| 31 |
+
{
|
| 32 |
+
"filters": {
|
| 33 |
+
"source_dataset": "ade20k"
|
| 34 |
+
},
|
| 35 |
+
"index_path": "<local-path>",
|
| 36 |
+
"name": "semseg_ade20k",
|
| 37 |
+
"num_rows": 20196,
|
| 38 |
+
"path": "<local-path>",
|
| 39 |
+
"sampling": "row",
|
| 40 |
+
"type": "indexed_prepared_tar",
|
| 41 |
+
"unique_data_fraction": 1.0,
|
| 42 |
+
"unique_data_seed": 42,
|
| 43 |
+
"weight": 0.04125
|
| 44 |
+
},
|
| 45 |
+
{
|
| 46 |
+
"filters": {
|
| 47 |
+
"source_dataset": "voc"
|
| 48 |
+
},
|
| 49 |
+
"index_path": "<local-path>",
|
| 50 |
+
"name": "semseg_voc",
|
| 51 |
+
"num_rows": 1464,
|
| 52 |
+
"path": "<local-path>",
|
| 53 |
+
"sampling": "row",
|
| 54 |
+
"type": "indexed_prepared_tar",
|
| 55 |
+
"unique_data_fraction": 1.0,
|
| 56 |
+
"unique_data_seed": 42,
|
| 57 |
+
"weight": 0.04125
|
| 58 |
+
},
|
| 59 |
+
{
|
| 60 |
+
"filters": {
|
| 61 |
+
"source_dataset": "ade20k_inst"
|
| 62 |
+
},
|
| 63 |
+
"index_path": "<local-path>",
|
| 64 |
+
"name": "instanceseg_ade20k",
|
| 65 |
+
"num_rows": 17277,
|
| 66 |
+
"path": "<local-path>",
|
| 67 |
+
"sampling": "row",
|
| 68 |
+
"type": "indexed_prepared_tar",
|
| 69 |
+
"unique_data_fraction": 1.0,
|
| 70 |
+
"unique_data_seed": 42,
|
| 71 |
+
"weight": 0.042
|
| 72 |
+
},
|
| 73 |
+
{
|
| 74 |
+
"filters": {
|
| 75 |
+
"source_dataset": "mapillary"
|
| 76 |
+
},
|
| 77 |
+
"index_path": "<local-path>",
|
| 78 |
+
"name": "instanceseg_mapillary",
|
| 79 |
+
"num_rows": 17966,
|
| 80 |
+
"path": "<local-path>",
|
| 81 |
+
"sampling": "row",
|
| 82 |
+
"type": "indexed_prepared_tar",
|
| 83 |
+
"unique_data_fraction": 1.0,
|
| 84 |
+
"unique_data_seed": 42,
|
| 85 |
+
"weight": 0.042
|
| 86 |
+
},
|
| 87 |
+
{
|
| 88 |
+
"filters": {
|
| 89 |
+
"source_dataset": "voc_inst"
|
| 90 |
+
},
|
| 91 |
+
"index_path": "<local-path>",
|
| 92 |
+
"name": "instanceseg_voc",
|
| 93 |
+
"num_rows": 1464,
|
| 94 |
+
"path": "<local-path>",
|
| 95 |
+
"sampling": "row",
|
| 96 |
+
"type": "indexed_prepared_tar",
|
| 97 |
+
"unique_data_fraction": 1.0,
|
| 98 |
+
"unique_data_seed": 42,
|
| 99 |
+
"weight": 0.021
|
| 100 |
+
},
|
| 101 |
+
{
|
| 102 |
+
"filters": null,
|
| 103 |
+
"index_path": "<local-path>",
|
| 104 |
+
"name": "refseg",
|
| 105 |
+
"num_rows": 50000,
|
| 106 |
+
"path": "<local-path>",
|
| 107 |
+
"sampling": "row",
|
| 108 |
+
"type": "indexed_prepared_tar",
|
| 109 |
+
"unique_data_fraction": 1.0,
|
| 110 |
+
"unique_data_seed": 42,
|
| 111 |
+
"weight": 0.105
|
| 112 |
+
},
|
| 113 |
+
{
|
| 114 |
+
"filters": null,
|
| 115 |
+
"index_path": "<local-path>",
|
| 116 |
+
"name": "pico_banana",
|
| 117 |
+
"num_rows": 257730,
|
| 118 |
+
"path": "<local-path>",
|
| 119 |
+
"sampling": "row",
|
| 120 |
+
"type": "indexed_prepared_tar",
|
| 121 |
+
"unique_data_fraction": 1.0,
|
| 122 |
+
"unique_data_seed": 42,
|
| 123 |
+
"weight": 0.165
|
| 124 |
+
},
|
| 125 |
+
{
|
| 126 |
+
"filters": null,
|
| 127 |
+
"index_path": "<local-path>",
|
| 128 |
+
"name": "identity",
|
| 129 |
+
"num_rows": 50000,
|
| 130 |
+
"path": "<local-path>",
|
| 131 |
+
"sampling": "row",
|
| 132 |
+
"type": "indexed_prepared_tar",
|
| 133 |
+
"unique_data_fraction": 1.0,
|
| 134 |
+
"unique_data_seed": 42,
|
| 135 |
+
"weight": 0.11
|
| 136 |
+
},
|
| 137 |
+
{
|
| 138 |
+
"filters": null,
|
| 139 |
+
"index_path": "<local-path>",
|
| 140 |
+
"name": "depth_hypersim",
|
| 141 |
+
"num_rows": 73814,
|
| 142 |
+
"path": "<local-path>",
|
| 143 |
+
"sampling": "scene_balanced",
|
| 144 |
+
"type": "indexed_prepared_tar",
|
| 145 |
+
"unique_data_fraction": 1.0,
|
| 146 |
+
"unique_data_seed": 42,
|
| 147 |
+
"weight": 0.045
|
| 148 |
+
},
|
| 149 |
+
{
|
| 150 |
+
"filters": null,
|
| 151 |
+
"index_path": "<local-path>",
|
| 152 |
+
"name": "depth_matterport3d",
|
| 153 |
+
"num_rows": 31838,
|
| 154 |
+
"path": "<local-path>",
|
| 155 |
+
"sampling": "scene_balanced",
|
| 156 |
+
"type": "indexed_prepared_tar",
|
| 157 |
+
"unique_data_fraction": 1.0,
|
| 158 |
+
"unique_data_seed": 42,
|
| 159 |
+
"weight": 0.045
|
| 160 |
+
},
|
| 161 |
+
{
|
| 162 |
+
"filters": null,
|
| 163 |
+
"index_path": "<local-path>",
|
| 164 |
+
"name": "depth_replica",
|
| 165 |
+
"num_rows": 9000,
|
| 166 |
+
"path": "<local-path>",
|
| 167 |
+
"sampling": "scene_balanced",
|
| 168 |
+
"type": "indexed_prepared_tar",
|
| 169 |
+
"unique_data_fraction": 1.0,
|
| 170 |
+
"unique_data_seed": 42,
|
| 171 |
+
"weight": 0.045
|
| 172 |
+
},
|
| 173 |
+
{
|
| 174 |
+
"filters": null,
|
| 175 |
+
"index_path": "<local-path>",
|
| 176 |
+
"name": "depth_synscapes",
|
| 177 |
+
"num_rows": 25000,
|
| 178 |
+
"path": "<local-path>",
|
| 179 |
+
"sampling": "row",
|
| 180 |
+
"type": "indexed_prepared_tar",
|
| 181 |
+
"unique_data_fraction": 1.0,
|
| 182 |
+
"unique_data_seed": 42,
|
| 183 |
+
"weight": 0.045
|
| 184 |
+
},
|
| 185 |
+
{
|
| 186 |
+
"filters": null,
|
| 187 |
+
"index_path": "<local-path>",
|
| 188 |
+
"name": "depth_taskonomy",
|
| 189 |
+
"num_rows": 24605,
|
| 190 |
+
"path": "<local-path>",
|
| 191 |
+
"sampling": "scene_balanced",
|
| 192 |
+
"type": "indexed_prepared_tar",
|
| 193 |
+
"unique_data_fraction": 1.0,
|
| 194 |
+
"unique_data_seed": 42,
|
| 195 |
+
"weight": 0.045
|
| 196 |
+
},
|
| 197 |
+
{
|
| 198 |
+
"filters": null,
|
| 199 |
+
"index_path": "<local-path>",
|
| 200 |
+
"name": "normals_hypersim",
|
| 201 |
+
"num_rows": 73794,
|
| 202 |
+
"path": "<local-path>",
|
| 203 |
+
"sampling": "scene_balanced",
|
| 204 |
+
"type": "indexed_prepared_tar",
|
| 205 |
+
"unique_data_fraction": 1.0,
|
| 206 |
+
"unique_data_seed": 42,
|
| 207 |
+
"weight": 0.025
|
| 208 |
+
},
|
| 209 |
+
{
|
| 210 |
+
"filters": null,
|
| 211 |
+
"index_path": "<local-path>",
|
| 212 |
+
"name": "normals_matterport3d",
|
| 213 |
+
"num_rows": 31838,
|
| 214 |
+
"path": "<local-path>",
|
| 215 |
+
"sampling": "scene_balanced",
|
| 216 |
+
"type": "indexed_prepared_tar",
|
| 217 |
+
"unique_data_fraction": 1.0,
|
| 218 |
+
"unique_data_seed": 42,
|
| 219 |
+
"weight": 0.025
|
| 220 |
+
},
|
| 221 |
+
{
|
| 222 |
+
"filters": null,
|
| 223 |
+
"index_path": "<local-path>",
|
| 224 |
+
"name": "normals_replica",
|
| 225 |
+
"num_rows": 9000,
|
| 226 |
+
"path": "<local-path>",
|
| 227 |
+
"sampling": "scene_balanced",
|
| 228 |
+
"type": "indexed_prepared_tar",
|
| 229 |
+
"unique_data_fraction": 1.0,
|
| 230 |
+
"unique_data_seed": 42,
|
| 231 |
+
"weight": 0.025
|
| 232 |
+
},
|
| 233 |
+
{
|
| 234 |
+
"filters": null,
|
| 235 |
+
"index_path": "<local-path>",
|
| 236 |
+
"name": "normals_synscapes",
|
| 237 |
+
"num_rows": 25000,
|
| 238 |
+
"path": "<local-path>",
|
| 239 |
+
"sampling": "row",
|
| 240 |
+
"type": "indexed_prepared_tar",
|
| 241 |
+
"unique_data_fraction": 1.0,
|
| 242 |
+
"unique_data_seed": 42,
|
| 243 |
+
"weight": 0.025
|
| 244 |
+
},
|
| 245 |
+
{
|
| 246 |
+
"filters": null,
|
| 247 |
+
"index_path": "<local-path>",
|
| 248 |
+
"name": "normals_taskonomy",
|
| 249 |
+
"num_rows": 24605,
|
| 250 |
+
"path": "<local-path>",
|
| 251 |
+
"sampling": "scene_balanced",
|
| 252 |
+
"type": "indexed_prepared_tar",
|
| 253 |
+
"unique_data_fraction": 1.0,
|
| 254 |
+
"unique_data_seed": 42,
|
| 255 |
+
"weight": 0.025
|
| 256 |
+
}
|
| 257 |
+
],
|
| 258 |
+
"samples_per_epoch": 160000,
|
| 259 |
+
"seed": 42,
|
| 260 |
+
"shuffle": true
|
| 261 |
+
}
|
ablation_models/additional_studies/cross_task/creative_task_fraction/generation0/checkpoint-50000/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/cross_task/creative_task_fraction/generation0/checkpoint-50000/run_spec.json
ADDED
|
@@ -0,0 +1,112 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"account": "h100_v115be14_mirror",
|
| 3 |
+
"accounts_sha256": "af3ceb90fa458653c105addccc12626788c0a29171542c0c153e8b333c357792",
|
| 4 |
+
"after_run": null,
|
| 5 |
+
"campaign_file": "<local-path>",
|
| 6 |
+
"campaign_id": "geometry_phase2_20260806_v1",
|
| 7 |
+
"campaign_sha256": "8d81de74c098f1b833f3e1f1e3bcc4c9cd1aef1079f74bf682b6fa08d6f7cd75",
|
| 8 |
+
"campaign_status": "optional_phase2",
|
| 9 |
+
"created_utc": "2026-08-11T10:23:26.231424+00:00",
|
| 10 |
+
"dataset_specs_sha256": "579119541be2828659f5295fea59e1690a12bf415c3f5f2bd55525304749fdd3",
|
| 11 |
+
"environment": {
|
| 12 |
+
"ALLOW_LEGACY_HYPERSIM_RANGE_DEPTH": "0",
|
| 13 |
+
"ALLOW_UNVERIFIED_NORMAL_SOURCES": "0",
|
| 14 |
+
"ANGULAR_NORMAL_AUX_WEIGHT": "0",
|
| 15 |
+
"ASPECT_RATIO_BUCKETS": "512,512;576,448;448,576;608,416;416,608",
|
| 16 |
+
"AUTO_RECONCILE_ON_COMPLETE": "0",
|
| 17 |
+
"CACHE_LATENTS": "0",
|
| 18 |
+
"CHECKPOINTING_STEPS": "1000",
|
| 19 |
+
"CHECKPOINTS_TOTAL_LIMIT": "3",
|
| 20 |
+
"CUDA_VISIBLE_DEVICES": "0,1,2,3",
|
| 21 |
+
"DATASET_POOL_SEED": "42",
|
| 22 |
+
"DATASET_SPECS": "<local-path>",
|
| 23 |
+
"DEPTH_ENCODING": "relative_inverse_grayscale",
|
| 24 |
+
"DEPTH_PROMPT_CONTRACT": "campaign_detailed_inverse",
|
| 25 |
+
"DEPTH_SOURCE_MULTIPLIERS": "{}",
|
| 26 |
+
"DEPTH_TARGET_RESIZE_POLICY": "metric_field_then_encode",
|
| 27 |
+
"DEPTH_WEIGHT": "0.225",
|
| 28 |
+
"DISTRIBUTED_STRATEGY": "fsdp",
|
| 29 |
+
"EVAL_GPU_GRES": "gpu:h100:4",
|
| 30 |
+
"EVAL_PARTITION": "h100",
|
| 31 |
+
"EVAL_QOS": "normal",
|
| 32 |
+
"EXCLUDE_DEPTH_SOURCE": "",
|
| 33 |
+
"EXCLUDE_NORMALS_SOURCE": "",
|
| 34 |
+
"FSDP_TRANSFORMER_DTYPE_POLICY": "fp32_all",
|
| 35 |
+
"GRADIENT_ACCUMULATION_STEPS": "2",
|
| 36 |
+
"GUIDANCE_SCALE": "3.5",
|
| 37 |
+
"IDENTITY_ARCHIVE": "<local-path>",
|
| 38 |
+
"IDENTITY_WEIGHT": "0.11",
|
| 39 |
+
"INSTANCE_WEIGHT": "0.105",
|
| 40 |
+
"LEARNING_RATE": "2e-6",
|
| 41 |
+
"LOSS_MASK_EROSION_RADIUS": "0",
|
| 42 |
+
"LOSS_MASK_MIN_SAMPLE_VALID_FRACTION": "0.0",
|
| 43 |
+
"LOSS_MASK_MIN_VALID_FRACTION": "0.5",
|
| 44 |
+
"LOSS_MASK_MODE": "hard",
|
| 45 |
+
"LOSS_MASK_POLICY": "auto",
|
| 46 |
+
"LOSS_MASK_REDUCTION": "per_sample",
|
| 47 |
+
"LR_SCHEDULER": "constant_with_warmup",
|
| 48 |
+
"LR_WARMUP_STEPS": "1000",
|
| 49 |
+
"MAX_TRAIN_STEPS": "50000",
|
| 50 |
+
"METRIC_DEPTH_AUX_WEIGHT": "0",
|
| 51 |
+
"MIXED_PRECISION": "bf16",
|
| 52 |
+
"MODEL": "<local-path>",
|
| 53 |
+
"MODEL_FAMILY": "flux2_klein",
|
| 54 |
+
"NORMALS_WEIGHT": "0.125",
|
| 55 |
+
"NORMAL_PROMPT_CONTRACT": "campaign_short",
|
| 56 |
+
"NORMAL_SOURCE_MULTIPLIERS": "{}",
|
| 57 |
+
"NORMAL_TARGET_ENCODING": "vision_banana_x_left_red",
|
| 58 |
+
"NORMAL_TARGET_RESIZE_POLICY": "vector_field_renormalize_then_encode",
|
| 59 |
+
"NUM_PROCESSES": "4",
|
| 60 |
+
"OUTPUT_DIR": "<local-path>",
|
| 61 |
+
"PALETTE_MANIFOLD_LOSS_WEIGHT": "0",
|
| 62 |
+
"PD12M_WEIGHT": "0.0",
|
| 63 |
+
"PERMANENT_CHECKPOINT_STEPS": "50000",
|
| 64 |
+
"PICO_ARCHIVE": "<local-path>",
|
| 65 |
+
"PICO_WEIGHT": "0.165",
|
| 66 |
+
"PROJECT_ROOT": "<local-path>",
|
| 67 |
+
"PROMPT_ENCODING_MODE": "auto",
|
| 68 |
+
"REFSEG_WEIGHT": "0.105",
|
| 69 |
+
"REPORT_TO": "tensorboard",
|
| 70 |
+
"RESAMPLE_DATASET_EACH_EPOCH": "1",
|
| 71 |
+
"RESOLUTION": "512",
|
| 72 |
+
"RESUME_FROM_CHECKPOINT": "latest",
|
| 73 |
+
"ROOT": "<local-path>",
|
| 74 |
+
"RUN_DIR": "<local-path>",
|
| 75 |
+
"RUN_ID": "p2_gen_fraction_0_s42",
|
| 76 |
+
"SAMPLES_PER_EPOCH": "160000",
|
| 77 |
+
"SEED": "42",
|
| 78 |
+
"SEMANTIC_WEIGHT": "0.165",
|
| 79 |
+
"SKIP_FINAL_INFERENCE": "1",
|
| 80 |
+
"SKIP_FINAL_SAVE": "1",
|
| 81 |
+
"TASK_HOMOGENEOUS_BATCHES": "1",
|
| 82 |
+
"TRAINING_SUITE": "geometry_phase2_20260806_v1",
|
| 83 |
+
"TRAIN_ALL_TRANSFORMER_BLOCKS": "1",
|
| 84 |
+
"TRAIN_BATCH_SIZE": "1",
|
| 85 |
+
"TRAIN_FULL_TRANSFORMER": "0",
|
| 86 |
+
"TRAIN_INPUT_LAYERS": "0",
|
| 87 |
+
"TRAIN_LAST_SINGLE_TRANSFORMER_BLOCKS": "0",
|
| 88 |
+
"TRAIN_LAST_TRANSFORMER_BLOCKS": "0",
|
| 89 |
+
"TRAIN_METHOD": "partial_full_rank",
|
| 90 |
+
"TRAIN_NON_BLOCK_LAYERS": "1",
|
| 91 |
+
"TRAIN_OUTPUT_LAYERS": "1",
|
| 92 |
+
"UNIT_NORMAL_LOSS_WEIGHT": "0",
|
| 93 |
+
"WEIGHTING_SCHEME": "none",
|
| 94 |
+
"WEIGHT_DECAY": "1e-4"
|
| 95 |
+
},
|
| 96 |
+
"family": "creative_fraction_sweep",
|
| 97 |
+
"full_eval_checkpoints": [
|
| 98 |
+
50000
|
| 99 |
+
],
|
| 100 |
+
"geometry_eval_checkpoints": [
|
| 101 |
+
50000
|
| 102 |
+
],
|
| 103 |
+
"geometry_gate": "<local-path>",
|
| 104 |
+
"geometry_gate_sha256": "762516ac340dca76c5fa3f9c45848823a32aa749582a5b2363cfaf9c6ee388eb",
|
| 105 |
+
"git_commit": "8d4f2df7242feabfe95039f2f3181894a0048c0e",
|
| 106 |
+
"group": "p2_creative_fraction_sweep",
|
| 107 |
+
"inputs_sha256": "4fab5be1229ad50bdbd21fd4406e864ae8f778f53308d7ae4df51f7e48615d4c",
|
| 108 |
+
"label": "Image generation 0% (0x default); equal-delta redistribution",
|
| 109 |
+
"run_id": "p2_gen_fraction_0_s42",
|
| 110 |
+
"schema_version": 1,
|
| 111 |
+
"status": "PLANNED"
|
| 112 |
+
}
|
ablation_models/additional_studies/cross_task/creative_task_fraction/generation0/checkpoint-50000/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": 1000,
|
| 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": 50000,
|
| 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": "50000",
|
| 82 |
+
"precompute_cache_dir": "<local-path>",
|
| 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 |
+
}
|