Upload ai-toolkit_example_job_config.json
Browse files
ai-toolkit_example_job_config.json
ADDED
|
@@ -0,0 +1,125 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"job": "extension",
|
| 3 |
+
"config": {
|
| 4 |
+
"name": "IdoInpaint_2",
|
| 5 |
+
"process": [
|
| 6 |
+
{
|
| 7 |
+
"type": "diffusion_trainer",
|
| 8 |
+
"training_folder": "/workspace/IdoInpaint_2/output",
|
| 9 |
+
"sqlite_db_path": "/workspace/ai-toolkit-ido4inpaintcdx/aitk_db.db",
|
| 10 |
+
"device": "cuda",
|
| 11 |
+
"trigger_word": null,
|
| 12 |
+
"performance_log_every": 10,
|
| 13 |
+
"network": {
|
| 14 |
+
"type": "lora",
|
| 15 |
+
"linear": 64,
|
| 16 |
+
"linear_alpha": 64,
|
| 17 |
+
"lokr_full_rank": true,
|
| 18 |
+
"lokr_factor": -1,
|
| 19 |
+
"network_kwargs": {
|
| 20 |
+
"ignore_if_contains": []
|
| 21 |
+
}
|
| 22 |
+
},
|
| 23 |
+
"save": {
|
| 24 |
+
"dtype": "bf16",
|
| 25 |
+
"save_every": 250,
|
| 26 |
+
"max_step_saves_to_keep": 25,
|
| 27 |
+
"save_format": "diffusers",
|
| 28 |
+
"push_to_hub": false
|
| 29 |
+
},
|
| 30 |
+
"datasets": [
|
| 31 |
+
{
|
| 32 |
+
"folder_path": "/workspace/dataset/IdoInpaint_2",
|
| 33 |
+
"mask_path": null,
|
| 34 |
+
"mask_min_value": 0.1,
|
| 35 |
+
"default_caption": "",
|
| 36 |
+
"caption_ext": "txt",
|
| 37 |
+
"caption_dropout_rate": 0.05,
|
| 38 |
+
"cache_latents_to_disk": true,
|
| 39 |
+
"is_reg": false,
|
| 40 |
+
"network_weight": 1,
|
| 41 |
+
"resolution": [
|
| 42 |
+
512,
|
| 43 |
+
1024,
|
| 44 |
+
1328
|
| 45 |
+
],
|
| 46 |
+
"controls": [],
|
| 47 |
+
"use_reference_images": true,
|
| 48 |
+
"shrink_video_to_frames": true,
|
| 49 |
+
"num_frames": 1,
|
| 50 |
+
"flip_x": false,
|
| 51 |
+
"flip_y": false,
|
| 52 |
+
"num_repeats": 1
|
| 53 |
+
}
|
| 54 |
+
],
|
| 55 |
+
"train": {
|
| 56 |
+
"batch_size": 1,
|
| 57 |
+
"bypass_guidance_embedding": false,
|
| 58 |
+
"steps": 7000,
|
| 59 |
+
"gradient_accumulation": 1,
|
| 60 |
+
"train_unet": true,
|
| 61 |
+
"train_text_encoder": false,
|
| 62 |
+
"gradient_checkpointing": true,
|
| 63 |
+
"noise_scheduler": "flowmatch",
|
| 64 |
+
"optimizer": "adamw8bit",
|
| 65 |
+
"timestep_type": "linear",
|
| 66 |
+
"content_or_style": "balanced",
|
| 67 |
+
"optimizer_params": {
|
| 68 |
+
"weight_decay": 0.0001
|
| 69 |
+
},
|
| 70 |
+
"unload_text_encoder": false,
|
| 71 |
+
"cache_text_embeddings": true,
|
| 72 |
+
"lr": 0.00005,
|
| 73 |
+
"ema_config": {
|
| 74 |
+
"use_ema": false,
|
| 75 |
+
"ema_decay": 0.99
|
| 76 |
+
},
|
| 77 |
+
"skip_first_sample": false,
|
| 78 |
+
"force_first_sample": false,
|
| 79 |
+
"disable_sampling": false,
|
| 80 |
+
"dtype": "bf16",
|
| 81 |
+
"diff_output_preservation": false,
|
| 82 |
+
"diff_output_preservation_multiplier": 1,
|
| 83 |
+
"diff_output_preservation_class": "person",
|
| 84 |
+
"switch_boundary_every": 1,
|
| 85 |
+
"loss_type": "mse"
|
| 86 |
+
},
|
| 87 |
+
"logging": {
|
| 88 |
+
"log_every": 1,
|
| 89 |
+
"use_ui_logger": true
|
| 90 |
+
},
|
| 91 |
+
"model": {
|
| 92 |
+
"name_or_path": "ideogram-ai/ideogram-4-fp8",
|
| 93 |
+
"quantize": true,
|
| 94 |
+
"qtype": "qfloat8",
|
| 95 |
+
"quantize_te": true,
|
| 96 |
+
"qtype_te": "qfloat8",
|
| 97 |
+
"arch": "ideogram4",
|
| 98 |
+
"low_vram": true,
|
| 99 |
+
"model_kwargs": {},
|
| 100 |
+
"layer_offloading": false,
|
| 101 |
+
"layer_offloading_text_encoder_percent": 1,
|
| 102 |
+
"layer_offloading_transformer_percent": 1
|
| 103 |
+
},
|
| 104 |
+
"sample": {
|
| 105 |
+
"sampler": "flowmatch",
|
| 106 |
+
"sample_every": 250,
|
| 107 |
+
"width": 1024,
|
| 108 |
+
"height": 1024,
|
| 109 |
+
"samples": [],
|
| 110 |
+
"neg": "",
|
| 111 |
+
"seed": 42,
|
| 112 |
+
"walk_seed": true,
|
| 113 |
+
"guidance_scale": 4,
|
| 114 |
+
"sample_steps": 30,
|
| 115 |
+
"num_frames": 1,
|
| 116 |
+
"fps": 1
|
| 117 |
+
}
|
| 118 |
+
}
|
| 119 |
+
]
|
| 120 |
+
},
|
| 121 |
+
"meta": {
|
| 122 |
+
"name": "[name]",
|
| 123 |
+
"version": "1.0"
|
| 124 |
+
}
|
| 125 |
+
}
|