Delete dit/f64_dit_t2m_score_50k/config.json
Browse files
dit/f64_dit_t2m_score_50k/config.json
DELETED
|
@@ -1,213 +0,0 @@
|
|
| 1 |
-
{
|
| 2 |
-
"resume": true,
|
| 3 |
-
"project_dir": "/mnt/pfs/users/boyuan.wang/exp/humandreamer/f64_dit_patchattn_t2m_score_s2_5w_fe2loss_open/",
|
| 4 |
-
"launch": {
|
| 5 |
-
"gpu_ids": [
|
| 6 |
-
3
|
| 7 |
-
],
|
| 8 |
-
"deepspeed_config": {
|
| 9 |
-
"gradient_accumulation_steps": 1,
|
| 10 |
-
"gradient_clipping": 1.0,
|
| 11 |
-
"offload_optimizer_device": "none",
|
| 12 |
-
"offload_param_device": "none",
|
| 13 |
-
"zero_stage": 2
|
| 14 |
-
}
|
| 15 |
-
},
|
| 16 |
-
"dataloaders": {
|
| 17 |
-
"train": {
|
| 18 |
-
"num_frames": 64,
|
| 19 |
-
"data_or_config": [
|
| 20 |
-
"/shared_disk/users/boyuan.wang/datasets/motionvid-release/Kinetics400_train",
|
| 21 |
-
"/shared_disk/users/boyuan.wang/datasets/motionvid-release/Kinetics400_test",
|
| 22 |
-
"/shared_disk/users/boyuan.wang/datasets/motionvid-release/Kinetics400_val",
|
| 23 |
-
"/shared_disk/users/boyuan.wang/datasets/motionvid-release/HMDB51",
|
| 24 |
-
"/shared_disk/users/boyuan.wang/datasets/motionvid-release/UBody",
|
| 25 |
-
"/shared_disk/users/boyuan.wang/datasets/motionvid-release/HAA500",
|
| 26 |
-
"/shared_disk/users/boyuan.wang/datasets/motionvid-release/DEFW",
|
| 27 |
-
"/shared_disk/users/boyuan.wang/datasets/motionvid-release/Charades",
|
| 28 |
-
"/shared_disk/users/boyuan.wang/datasets/motionvid-release/CharadesEgo",
|
| 29 |
-
"/shared_disk/users/boyuan.wang/datasets/motionvid-release/CAER",
|
| 30 |
-
"/shared_disk/users/boyuan.wang/datasets/motionvid-release/Something-Something-v2",
|
| 31 |
-
"/shared_disk/users/boyuan.wang/datasets/motionvid-release/Kinetics700",
|
| 32 |
-
"/shared_disk/users/boyuan.wang/datasets/motionvid-release/Motion-X",
|
| 33 |
-
"/shared_disk/users/boyuan.wang/datasets/motionvid-release/ActivityNet200",
|
| 34 |
-
"/shared_disk/users/boyuan.wang/datasets/motionvid-release/Panda/chunk1",
|
| 35 |
-
"/shared_disk/users/boyuan.wang/datasets/motionvid-release/Panda/chunk2",
|
| 36 |
-
"/shared_disk/users/boyuan.wang/datasets/motionvid-release/Panda/chunk3",
|
| 37 |
-
"/shared_disk/users/boyuan.wang/datasets/motionvid-release/Panda/chunk4",
|
| 38 |
-
"/shared_disk/users/boyuan.wang/datasets/motionvid-release/Panda/chunk5",
|
| 39 |
-
"/shared_disk/users/boyuan.wang/datasets/motionvid-release/Panda/chunk6",
|
| 40 |
-
"/shared_disk/users/boyuan.wang/datasets/motionvid-release/Panda/chunk7",
|
| 41 |
-
"/shared_disk/users/boyuan.wang/datasets/motionvid-release/Panda/chunk8",
|
| 42 |
-
"/shared_disk/users/boyuan.wang/datasets/motionvid-release/Panda/chunk9",
|
| 43 |
-
"/shared_disk/users/boyuan.wang/datasets/motionvid-release/Internet/chunk1",
|
| 44 |
-
"/shared_disk/users/boyuan.wang/datasets/motionvid-release/Internet/chunk2",
|
| 45 |
-
"/shared_disk/users/boyuan.wang/datasets/motionvid-release/Internet/chunk3",
|
| 46 |
-
"/shared_disk/users/boyuan.wang/datasets/motionvid-release/Internet/chunk4",
|
| 47 |
-
"/shared_disk/users/boyuan.wang/datasets/motionvid-release/Internet/chunk5",
|
| 48 |
-
"/shared_disk/users/boyuan.wang/datasets/motionvid-release/Internet/chunk6",
|
| 49 |
-
"/shared_disk/users/boyuan.wang/datasets/motionvid-release/Internet/chunk7",
|
| 50 |
-
"/shared_disk/users/boyuan.wang/datasets/motionvid-release/Internet/chunk8",
|
| 51 |
-
"/shared_disk/users/boyuan.wang/datasets/motionvid-release/Internet/chunk9",
|
| 52 |
-
"/shared_disk/users/boyuan.wang/datasets/motionvid-release/Internet/chunk10",
|
| 53 |
-
"/shared_disk/users/boyuan.wang/datasets/motionvid-release/Internet/chunk11",
|
| 54 |
-
"/shared_disk/users/boyuan.wang/datasets/motionvid-release/Internet/chunk12",
|
| 55 |
-
"/shared_disk/users/boyuan.wang/datasets/motionvid-release/Internet/chunk13",
|
| 56 |
-
"/shared_disk/users/boyuan.wang/datasets/motionvid-release/Internet/chunk14",
|
| 57 |
-
"/shared_disk/users/boyuan.wang/datasets/motionvid-release/Internet/chunk15",
|
| 58 |
-
"/shared_disk/users/boyuan.wang/datasets/motionvid-release/Internet/chunk16",
|
| 59 |
-
"/shared_disk/users/boyuan.wang/datasets/motionvid-release/Internet/chunk17"
|
| 60 |
-
],
|
| 61 |
-
"batch_size_per_gpu": 1,
|
| 62 |
-
"num_workers": 1,
|
| 63 |
-
"sampler": {
|
| 64 |
-
"type": "TrainTestSampler",
|
| 65 |
-
"stage": "train",
|
| 66 |
-
"percentage": 0.999
|
| 67 |
-
},
|
| 68 |
-
"transform": {
|
| 69 |
-
"type": "DITTransform",
|
| 70 |
-
"num_frames": 64,
|
| 71 |
-
"is_train": true,
|
| 72 |
-
"num_joints": 18,
|
| 73 |
-
"train_mode": "dit",
|
| 74 |
-
"mask_type": "score",
|
| 75 |
-
"stride": 2
|
| 76 |
-
},
|
| 77 |
-
"ref_frames": null
|
| 78 |
-
},
|
| 79 |
-
"test": {
|
| 80 |
-
"data_or_config": [
|
| 81 |
-
"/shared_disk/users/boyuan.wang/datasets/motionvid-release/Kinetics400_train",
|
| 82 |
-
"/shared_disk/users/boyuan.wang/datasets/motionvid-release/Kinetics400_test",
|
| 83 |
-
"/shared_disk/users/boyuan.wang/datasets/motionvid-release/Kinetics400_val",
|
| 84 |
-
"/shared_disk/users/boyuan.wang/datasets/motionvid-release/HMDB51",
|
| 85 |
-
"/shared_disk/users/boyuan.wang/datasets/motionvid-release/UBody",
|
| 86 |
-
"/shared_disk/users/boyuan.wang/datasets/motionvid-release/HAA500",
|
| 87 |
-
"/shared_disk/users/boyuan.wang/datasets/motionvid-release/DEFW",
|
| 88 |
-
"/shared_disk/users/boyuan.wang/datasets/motionvid-release/Charades",
|
| 89 |
-
"/shared_disk/users/boyuan.wang/datasets/motionvid-release/CharadesEgo"
|
| 90 |
-
],
|
| 91 |
-
"frame_num": 64,
|
| 92 |
-
"batch_size_per_gpu": 1,
|
| 93 |
-
"num_workers": 1,
|
| 94 |
-
"sampler": {
|
| 95 |
-
"type": "TrainTestSampler",
|
| 96 |
-
"stage": "test",
|
| 97 |
-
"percentage": 0.001
|
| 98 |
-
},
|
| 99 |
-
"transform": {
|
| 100 |
-
"type": "DITTransform",
|
| 101 |
-
"num_frames": 64,
|
| 102 |
-
"is_train": false,
|
| 103 |
-
"num_joints": 18,
|
| 104 |
-
"train_mode": "dit",
|
| 105 |
-
"mask_type": "score",
|
| 106 |
-
"stride": 2
|
| 107 |
-
},
|
| 108 |
-
"ref_frames": null
|
| 109 |
-
}
|
| 110 |
-
},
|
| 111 |
-
"models": {
|
| 112 |
-
"pose_dim": 256,
|
| 113 |
-
"gen_type": "t2m",
|
| 114 |
-
"use_pose_attention": false,
|
| 115 |
-
"train_mode": "dit",
|
| 116 |
-
"mask_type": "score",
|
| 117 |
-
"loss_type": "fe2",
|
| 118 |
-
"vae": {
|
| 119 |
-
"pretrained": null,
|
| 120 |
-
"dtype": "float32",
|
| 121 |
-
"hyperparameters": {
|
| 122 |
-
"in_channels": 3,
|
| 123 |
-
"out_channels": 3,
|
| 124 |
-
"down_block_types": [
|
| 125 |
-
"DownBlock1D",
|
| 126 |
-
"DownBlock1D",
|
| 127 |
-
"DownBlock1D",
|
| 128 |
-
"DownBlock1D"
|
| 129 |
-
],
|
| 130 |
-
"block_out_channels": [
|
| 131 |
-
128,
|
| 132 |
-
256,
|
| 133 |
-
512,
|
| 134 |
-
512
|
| 135 |
-
],
|
| 136 |
-
"layers_per_block": 2,
|
| 137 |
-
"latent_channels": 4,
|
| 138 |
-
"sample_size": 768
|
| 139 |
-
}
|
| 140 |
-
},
|
| 141 |
-
"dit": {
|
| 142 |
-
"dytpe": "float32",
|
| 143 |
-
"hyperparameters": {
|
| 144 |
-
"block_out_channels": [
|
| 145 |
-
"__tuple__",
|
| 146 |
-
320,
|
| 147 |
-
640,
|
| 148 |
-
1280
|
| 149 |
-
],
|
| 150 |
-
"addition_time_embed_dim": 256,
|
| 151 |
-
"num_attention_heads": [
|
| 152 |
-
"__tuple__",
|
| 153 |
-
5,
|
| 154 |
-
10,
|
| 155 |
-
20
|
| 156 |
-
],
|
| 157 |
-
"projection_class_embeddings_input_dim": 768,
|
| 158 |
-
"transformer_layers_per_block": 1,
|
| 159 |
-
"layers_per_block": 2,
|
| 160 |
-
"sample_size": 96,
|
| 161 |
-
"in_channels": 4,
|
| 162 |
-
"out_channels": 4,
|
| 163 |
-
"down_block_types": [
|
| 164 |
-
"__tuple__",
|
| 165 |
-
"CrossAttnDownBlockSpatioTemporal1D_nosample",
|
| 166 |
-
"CrossAttnDownBlockSpatioTemporal1D_nosample",
|
| 167 |
-
"DownBlockSpatioTemporal1D_nosample"
|
| 168 |
-
],
|
| 169 |
-
"up_block_types": [
|
| 170 |
-
"__tuple__",
|
| 171 |
-
"UpBlockSpatioTemporal1D_nosample",
|
| 172 |
-
"CrossAttnUpBlockSpatioTemporal1D_nosample",
|
| 173 |
-
"CrossAttnUpBlockSpatioTemporal1D_nosample"
|
| 174 |
-
],
|
| 175 |
-
"cross_attention_dim": 1024,
|
| 176 |
-
"num_frames": 64,
|
| 177 |
-
"loss_type": "fe2",
|
| 178 |
-
"patch_size": 2,
|
| 179 |
-
"need_attn": true
|
| 180 |
-
}
|
| 181 |
-
},
|
| 182 |
-
"vae_weight_path": "/shared_disk/users/boyuan.wang/exp/mdm/mdm_f64_k400_wholebody_predscore_vae1d_kl/models/checkpoint_epoch_400_step_193200/vae/diffusion_pytorch_model.bin",
|
| 183 |
-
"dit_weight_path": "/mnt/pfs/users/boyuan.wang/exp/humandreamer/f64_dit_patchattn_t2m_score_s2_5w_fe2loss_open/models/ema/dit"
|
| 184 |
-
},
|
| 185 |
-
"optimizers": {
|
| 186 |
-
"type": "AdamW",
|
| 187 |
-
"lr": 1e-05,
|
| 188 |
-
"weight_decay": 0.01
|
| 189 |
-
},
|
| 190 |
-
"schedulers": {
|
| 191 |
-
"name": "constant",
|
| 192 |
-
"num_warmup_steps": 0
|
| 193 |
-
},
|
| 194 |
-
"train": {
|
| 195 |
-
"max_epochs": 200,
|
| 196 |
-
"gradient_accumulation_steps": 1,
|
| 197 |
-
"mixed_precision": "no",
|
| 198 |
-
"checkpoint_interval": 10,
|
| 199 |
-
"checkpoint_total_limit": 10,
|
| 200 |
-
"log_with": "tensorboard",
|
| 201 |
-
"log_interval": 50,
|
| 202 |
-
"max_grad_norm": 1.0,
|
| 203 |
-
"activation_checkpointing": true,
|
| 204 |
-
"save_ema": true,
|
| 205 |
-
"percentage": 0.03968253968253968
|
| 206 |
-
},
|
| 207 |
-
"test": {
|
| 208 |
-
"mixed_precision": "fp16",
|
| 209 |
-
"vae_save_dir": "/mnt/pfs/users/boyuan.wang/exp/humandreamer/f64_dit_patchattn_t2m_score_s2_5w_fe2loss_open/vae_vis/checkpoint_epoch_400_step_193200",
|
| 210 |
-
"dit_save_dir": "/mnt/pfs/users/boyuan.wang/exp/humandreamer/f64_dit_patchattn_t2m_score_s2_5w_fe2loss_open/dit_vis/checkpoint_epoch_20_step_20",
|
| 211 |
-
"num_joints": 18
|
| 212 |
-
}
|
| 213 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|