Upload config.json with huggingface_hub
Browse files- config.json +107 -0
config.json
ADDED
|
@@ -0,0 +1,107 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"model": {
|
| 3 |
+
"base_learning_rate": 8e-05,
|
| 4 |
+
"target": "ldm.models.diffusion.ddpm.LatentDiffusion",
|
| 5 |
+
"params": {
|
| 6 |
+
"linear_start": 0.0015,
|
| 7 |
+
"linear_end": 0.0195,
|
| 8 |
+
"num_timesteps_cond": 1,
|
| 9 |
+
"log_every_t": 200,
|
| 10 |
+
"timesteps": 1000,
|
| 11 |
+
"first_stage_key": "image",
|
| 12 |
+
"cond_stage_key": "caption",
|
| 13 |
+
"hybrid_key": "c_concat",
|
| 14 |
+
"image_size": 64,
|
| 15 |
+
"channels": 3,
|
| 16 |
+
"cond_stage_trainable": true,
|
| 17 |
+
"conditioning_key": "hybrid",
|
| 18 |
+
"monitor": "val/loss_simple_ema",
|
| 19 |
+
"unet_config": {
|
| 20 |
+
"target": "ldm.modules.diffusionmodules.openaimodel.UNetModel",
|
| 21 |
+
"params": {
|
| 22 |
+
"image_size": 64,
|
| 23 |
+
"in_channels": 24,
|
| 24 |
+
"out_channels": 3,
|
| 25 |
+
"model_channels": 192,
|
| 26 |
+
"attention_resolutions": [
|
| 27 |
+
8,
|
| 28 |
+
4,
|
| 29 |
+
2
|
| 30 |
+
],
|
| 31 |
+
"num_res_blocks": 2,
|
| 32 |
+
"channel_mult": [
|
| 33 |
+
1,
|
| 34 |
+
2,
|
| 35 |
+
3,
|
| 36 |
+
5
|
| 37 |
+
],
|
| 38 |
+
"num_head_channels": 32,
|
| 39 |
+
"use_spatial_transformer": true,
|
| 40 |
+
"transformer_depth": 1,
|
| 41 |
+
"context_dim": 640
|
| 42 |
+
}
|
| 43 |
+
},
|
| 44 |
+
"first_stage_config": {
|
| 45 |
+
"target": "ldm.models.autoencoder.VQModelInterface",
|
| 46 |
+
"params": {
|
| 47 |
+
"embed_dim": 3,
|
| 48 |
+
"n_embed": 8192,
|
| 49 |
+
"monitor": "val/rec_loss",
|
| 50 |
+
"ddconfig": {
|
| 51 |
+
"double_z": false,
|
| 52 |
+
"z_channels": 3,
|
| 53 |
+
"resolution": 256,
|
| 54 |
+
"in_channels": 3,
|
| 55 |
+
"out_ch": 3,
|
| 56 |
+
"ch": 128,
|
| 57 |
+
"ch_mult": [
|
| 58 |
+
1,
|
| 59 |
+
2,
|
| 60 |
+
4
|
| 61 |
+
],
|
| 62 |
+
"num_res_blocks": 2,
|
| 63 |
+
"attn_resolutions": [],
|
| 64 |
+
"dropout": 0.0
|
| 65 |
+
},
|
| 66 |
+
"lossconfig": {
|
| 67 |
+
"target": "torch.nn.Identity"
|
| 68 |
+
}
|
| 69 |
+
}
|
| 70 |
+
},
|
| 71 |
+
"cond_stage_config": {
|
| 72 |
+
"target": "ldm.modules.encoders.modules.BERTEmbedder",
|
| 73 |
+
"params": {
|
| 74 |
+
"n_embed": 640,
|
| 75 |
+
"n_layer": 32
|
| 76 |
+
}
|
| 77 |
+
}
|
| 78 |
+
}
|
| 79 |
+
},
|
| 80 |
+
"data": {
|
| 81 |
+
"target": "data.datasets.CsllmTrainSeq",
|
| 82 |
+
"params": {
|
| 83 |
+
"batch_size": 8,
|
| 84 |
+
"num_workers": 12,
|
| 85 |
+
"wrap": false,
|
| 86 |
+
"train": {
|
| 87 |
+
"target": "data.datasets.CsllmTrainSeq",
|
| 88 |
+
"params": {
|
| 89 |
+
"config": {
|
| 90 |
+
"size": 256
|
| 91 |
+
}
|
| 92 |
+
}
|
| 93 |
+
}
|
| 94 |
+
}
|
| 95 |
+
},
|
| 96 |
+
"lightning": {
|
| 97 |
+
"trainer": {
|
| 98 |
+
"benchmark": false,
|
| 99 |
+
"max_epochs": 200,
|
| 100 |
+
"accelerator": "gpu",
|
| 101 |
+
"gpus": 1,
|
| 102 |
+
"gradient_clip_val": 1,
|
| 103 |
+
"checkpoint_callback": false,
|
| 104 |
+
"callbacks": []
|
| 105 |
+
}
|
| 106 |
+
}
|
| 107 |
+
}
|