AaronCIH commited on
Commit
12cf6e0
·
verified ·
1 Parent(s): 44765a7

Upload folder using huggingface_hub

Browse files
Files changed (48) hide show
  1. .gitattributes +2 -0
  2. sd35m_d2c_breif/checkpoints/epoch_100_step_302601.pth.pth +3 -0
  3. sd35m_d2c_breif/config.yaml +221 -0
  4. sd35m_d2c_breif/metric_helper/epoch_100_step_302601.pth.txt +2 -0
  5. sd35m_d2c_breif/metric_helper/epoch_10_step_30261.pth.txt +2 -0
  6. sd35m_d2c_breif/metric_helper/epoch_15_step_45391.pth.txt +2 -0
  7. sd35m_d2c_breif/metric_helper/epoch_20_step_60521.pth.txt +2 -0
  8. sd35m_d2c_breif/metric_helper/epoch_25_step_75651.pth.txt +2 -0
  9. sd35m_d2c_breif/metric_helper/epoch_30_step_90781.pth.txt +2 -0
  10. sd35m_d2c_breif/metric_helper/epoch_35_step_105911.pth.txt +2 -0
  11. sd35m_d2c_breif/metric_helper/epoch_40_step_121041.pth.txt +2 -0
  12. sd35m_d2c_breif/metric_helper/epoch_45_step_136171.pth.txt +2 -0
  13. sd35m_d2c_breif/metric_helper/epoch_50_step_151301.pth.txt +2 -0
  14. sd35m_d2c_breif/metric_helper/epoch_55_step_166431.pth.txt +2 -0
  15. sd35m_d2c_breif/metric_helper/epoch_5_step_15131.pth.txt +2 -0
  16. sd35m_d2c_breif/metric_helper/epoch_60_step_181561.pth.txt +2 -0
  17. sd35m_d2c_breif/metric_helper/epoch_65_step_196691.pth.txt +2 -0
  18. sd35m_d2c_breif/metric_helper/epoch_70_step_211821.pth.txt +2 -0
  19. sd35m_d2c_breif/metric_helper/epoch_75_step_226951.pth.txt +2 -0
  20. sd35m_d2c_breif/metric_helper/epoch_80_step_242081.pth.txt +2 -0
  21. sd35m_d2c_breif/metric_helper/epoch_85_step_257211.pth.txt +2 -0
  22. sd35m_d2c_breif/metric_helper/epoch_90_step_272341.pth.txt +2 -0
  23. sd35m_d2c_breif/metric_helper/epoch_95_step_287471.pth.txt +2 -0
  24. sd35m_d2c_breif/train_log.log +3 -0
  25. sd35m_p2p_breif/checkpoints/epoch_100_step_302601.pth +3 -0
  26. sd35m_p2p_breif/checkpoints/latest.pth +3 -0
  27. sd35m_p2p_breif/config.yaml +221 -0
  28. sd35m_p2p_breif/metric_helper/epoch_100_step_302601.pth.txt +2 -0
  29. sd35m_p2p_breif/metric_helper/epoch_10_step_30261.pth.txt +2 -0
  30. sd35m_p2p_breif/metric_helper/epoch_15_step_45391.pth.txt +2 -0
  31. sd35m_p2p_breif/metric_helper/epoch_20_step_60521.pth.txt +2 -0
  32. sd35m_p2p_breif/metric_helper/epoch_25_step_75651.pth.txt +2 -0
  33. sd35m_p2p_breif/metric_helper/epoch_30_step_90781.pth.txt +2 -0
  34. sd35m_p2p_breif/metric_helper/epoch_35_step_105911.pth.txt +2 -0
  35. sd35m_p2p_breif/metric_helper/epoch_40_step_121041.pth.txt +2 -0
  36. sd35m_p2p_breif/metric_helper/epoch_45_step_136171.pth.txt +2 -0
  37. sd35m_p2p_breif/metric_helper/epoch_50_step_151301.pth.txt +2 -0
  38. sd35m_p2p_breif/metric_helper/epoch_55_step_166431.pth.txt +2 -0
  39. sd35m_p2p_breif/metric_helper/epoch_5_step_15131.pth.txt +2 -0
  40. sd35m_p2p_breif/metric_helper/epoch_60_step_181561.pth.txt +2 -0
  41. sd35m_p2p_breif/metric_helper/epoch_65_step_196691.pth.txt +2 -0
  42. sd35m_p2p_breif/metric_helper/epoch_70_step_211821.pth.txt +2 -0
  43. sd35m_p2p_breif/metric_helper/epoch_75_step_226951.pth.txt +2 -0
  44. sd35m_p2p_breif/metric_helper/epoch_80_step_242081.pth.txt +2 -0
  45. sd35m_p2p_breif/metric_helper/epoch_85_step_257211.pth.txt +2 -0
  46. sd35m_p2p_breif/metric_helper/epoch_90_step_272341.pth.txt +2 -0
  47. sd35m_p2p_breif/metric_helper/epoch_95_step_287471.pth.txt +2 -0
  48. sd35m_p2p_breif/train_log.log +3 -0
.gitattributes CHANGED
@@ -33,3 +33,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
 
 
 
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
+ sd35m_d2c_breif/train_log.log filter=lfs diff=lfs merge=lfs -text
37
+ sd35m_p2p_breif/train_log.log filter=lfs diff=lfs merge=lfs -text
sd35m_d2c_breif/checkpoints/epoch_100_step_302601.pth.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2d3245d1dd8e3deb9d38b912871ec15283f7bb29c0be4823e1ea4fe7abaed43b
3
+ size 13933690163
sd35m_d2c_breif/config.yaml ADDED
@@ -0,0 +1,221 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ data:
2
+ data_dir:
3
+ - /home/work/shared-fi-datasets-01/users/hsiang.chen/Project/Datasets/IR
4
+ caption_proportion:
5
+ prompt: 1
6
+ external_caption_suffixes: []
7
+ external_clipscore_suffixes: []
8
+ clip_thr_temperature: 0.1
9
+ clip_thr: 25.0
10
+ del_img_clip_thr: 0.0
11
+ sort_dataset: false
12
+ load_text_feat: false
13
+ load_vae_feat: false
14
+ transform: default_train
15
+ type: IRImgDataset
16
+ image_size: 256
17
+ hq_only: false
18
+ valid_num: 0
19
+ data: null
20
+ extra: null
21
+ dset: train_brief
22
+ max_samples: null
23
+ model:
24
+ model: SD35M_D2C
25
+ model_pretrained: ./checkpoints/stable-diffusion-3.5-medium/sd3.5_medium.safetensors
26
+ shift: 3.0
27
+ teacher: null
28
+ input_channel: 16
29
+ image_size: 256
30
+ mixed_precision: bf16
31
+ fp32_attention: true
32
+ load_from: null
33
+ discriminator_model: null
34
+ teacher_model: null
35
+ teacher_model_weight_dtype: null
36
+ resume_from:
37
+ checkpoint: latest
38
+ load_ema: false
39
+ resume_optimizer: true
40
+ resume_lr_scheduler: true
41
+ aspect_ratio_type: ASPECT_RATIO_1024
42
+ multi_scale: false
43
+ pe_interpolation: 1.0
44
+ micro_condition: false
45
+ attn_type: linear
46
+ autocast_linear_attn: false
47
+ ffn_type: glumbconv
48
+ mlp_acts:
49
+ - silu
50
+ - silu
51
+ - null
52
+ mlp_ratio: 2.5
53
+ use_pe: false
54
+ pos_embed_type: sincos
55
+ qk_norm: false
56
+ class_dropout_prob: 0.1
57
+ linear_head_dim: 32
58
+ cross_norm: false
59
+ cross_attn_type: flash
60
+ logvar: false
61
+ cfg_scale: 4
62
+ cfg_embed: false
63
+ cfg_embed_scale: 1.0
64
+ guidance_type: classifier-free
65
+ pag_applied_layers:
66
+ - 8
67
+ ladd_multi_scale: true
68
+ head_block_ids: null
69
+ extra: null
70
+ vae:
71
+ vae_type: SDVAE
72
+ vae_pretrained: ./checkpoints/stable-diffusion-3.5-medium/sd3.5_medium.safetensors
73
+ weight_dtype: float32
74
+ scale_factor: 0.41407
75
+ vae_latent_dim: 16
76
+ vae_downsample_rate: 8
77
+ sample_posterior: true
78
+ extra: null
79
+ text_encoder:
80
+ text_encoder_name: sd35-text
81
+ text_encoder_pretrained: ./checkpoints/stable-diffusion-3.5-medium/text_encoders
82
+ caption_channels: 4096
83
+ y_norm: true
84
+ y_norm_scale_factor: 0.01
85
+ model_max_length: 300
86
+ chi_prompt:
87
+ - a photo of a cat
88
+ - Convenience store entrance at night. On the glass door, a vinyl decal reads
89
+ 'OPEN FOR QUALITY'. Inside, shelves and fluorescent lights; outside, a cyclist
90
+ passing by
91
+ - Sunrise beach, shallow tide washing over smooth sand. A piece of weathered driftwood
92
+ lies near the shoreline with a subtle branded text [SOS] on its surface; wet
93
+ sand reflections, micro-ripples, sun flare at horizon.
94
+ extra: null
95
+ scheduler:
96
+ train_sampling_steps: 1000
97
+ predict_flow_v: true
98
+ noise_schedule: linear_flow
99
+ pred_sigma: false
100
+ learn_sigma: true
101
+ vis_sampler: flow_dpm-solver
102
+ flow_shift: 3.0
103
+ weighting_scheme: logit_normal
104
+ weighting_scheme_discriminator: logit_normal_trigflow
105
+ add_noise_timesteps:
106
+ - 1.5708
107
+ logit_mean: 0.0
108
+ logit_std: 1.0
109
+ logit_mean_discriminator: 0.0
110
+ logit_std_discriminator: 1.0
111
+ sigma_data: 0.5
112
+ timestep_norm_scale_factor: 1.0
113
+ extra: null
114
+ train:
115
+ num_workers: 10
116
+ seed: 1229
117
+ train_batch_size: 4
118
+ num_epochs: 100
119
+ gradient_accumulation_steps: 8
120
+ grad_checkpointing: true
121
+ gradient_clip: 0.1
122
+ gc_step: 1
123
+ optimizer:
124
+ betas:
125
+ - 0.9
126
+ - 0.999
127
+ - 0.9999
128
+ eps:
129
+ - 1.0e-30
130
+ - 1.0e-16
131
+ lr: 5.0e-05
132
+ type: CAMEWrapper
133
+ weight_decay: 0.0
134
+ optimizer_D:
135
+ eps: 1.0e-10
136
+ lr: 0.0001
137
+ type: AdamW
138
+ weight_decay: 0.03
139
+ load_from_optimizer: false
140
+ load_from_lr_scheduler: false
141
+ resume_lr_scheduler: true
142
+ lr_schedule: cosine
143
+ lr_schedule_args:
144
+ num_warmup_steps: 2000
145
+ auto_lr:
146
+ rule: sqrt
147
+ eval_batch_size: 16
148
+ use_fsdp: false
149
+ use_flash_attn: false
150
+ eval_sampling_steps: 500
151
+ lora_rank: 4
152
+ log_interval: 1
153
+ mask_type: 'null'
154
+ mask_loss_coef: 0.0
155
+ load_mask_index: false
156
+ snr_loss: false
157
+ real_prompt_ratio: 1.0
158
+ early_stop_hours: 10000.0
159
+ save_image_epochs: 1
160
+ save_model_epochs: 5
161
+ save_model_steps: 500
162
+ visualize: true
163
+ null_embed_root: output/pretrained_models/
164
+ valid_prompt_embed_root: output/tmp_embed/
165
+ validation_prompts:
166
+ - dog
167
+ - portrait photo of a girl, photograph, highly detailed face, depth of field
168
+ - Self-portrait oil painting, a beautiful cyborg with golden hair, 8k
169
+ - Astronaut in a jungle, cold color palette, muted colors, detailed, 8k
170
+ - A photo of beautiful mountain with realistic sunset and blue lake, highly detailed,
171
+ masterpiece
172
+ local_save_vis: true
173
+ deterministic_validation: true
174
+ online_metric: false
175
+ eval_metric_step: 2000
176
+ online_metric_dir: metric_helper
177
+ work_dir: output/sd35m_d2c
178
+ skip_step: 0
179
+ loss_type: huber
180
+ huber_c: 0.001
181
+ num_ddim_timesteps: 50
182
+ ema_decay: 0.95
183
+ debug_nan: false
184
+ ema_update: false
185
+ ema_rate: 0.9999
186
+ tangent_warmup_steps: 10000
187
+ scm_cfg_scale:
188
+ - 1.0
189
+ cfg_interval: null
190
+ scm_logvar_loss: true
191
+ norm_invariant_to_spatial_dim: true
192
+ norm_same_as_512_scale: false
193
+ g_norm_constant: 0.1
194
+ g_norm_r: 1.0
195
+ show_gradient: false
196
+ lr_scale: null
197
+ adv_lambda: 1.0
198
+ scm_loss: true
199
+ scm_lambda: 1.0
200
+ loss_scale: 1.0
201
+ r1_penalty: false
202
+ r1_penalty_weight: 1.0e-05
203
+ diff_timesteps_D: true
204
+ suffix_checkpoints: disc
205
+ misaligned_pairs_D: false
206
+ discriminator_loss: cross entropy
207
+ largest_timestep: 1.5708
208
+ train_largest_timestep: false
209
+ largest_timestep_prob: 0.5
210
+ extra: null
211
+ controlnet: null
212
+ model_growth: null
213
+ work_dir: output/sd35m_d2c_breif
214
+ resume_from: latest
215
+ load_from: null
216
+ debug: true
217
+ caching: false
218
+ report_to: tensorboard
219
+ tracker_project_name: sana-baseline
220
+ name: tmp
221
+ loss_report_name: loss
sd35m_d2c_breif/metric_helper/epoch_100_step_302601.pth.txt ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ output/sd35m_d2c_breif/config.py
2
+ output/sd35m_d2c_breif/checkpoints/epoch_100_step_302601.pth
sd35m_d2c_breif/metric_helper/epoch_10_step_30261.pth.txt ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ output/sd35m_d2c_breif/config.py
2
+ output/sd35m_d2c_breif/checkpoints/epoch_10_step_30261.pth
sd35m_d2c_breif/metric_helper/epoch_15_step_45391.pth.txt ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ output/sd35m_d2c_breif/config.py
2
+ output/sd35m_d2c_breif/checkpoints/epoch_15_step_45391.pth
sd35m_d2c_breif/metric_helper/epoch_20_step_60521.pth.txt ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ output/sd35m_d2c_breif/config.py
2
+ output/sd35m_d2c_breif/checkpoints/epoch_20_step_60521.pth
sd35m_d2c_breif/metric_helper/epoch_25_step_75651.pth.txt ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ output/sd35m_d2c_breif/config.py
2
+ output/sd35m_d2c_breif/checkpoints/epoch_25_step_75651.pth
sd35m_d2c_breif/metric_helper/epoch_30_step_90781.pth.txt ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ output/sd35m_d2c_breif/config.py
2
+ output/sd35m_d2c_breif/checkpoints/epoch_30_step_90781.pth
sd35m_d2c_breif/metric_helper/epoch_35_step_105911.pth.txt ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ output/sd35m_d2c_breif/config.py
2
+ output/sd35m_d2c_breif/checkpoints/epoch_35_step_105911.pth
sd35m_d2c_breif/metric_helper/epoch_40_step_121041.pth.txt ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ output/sd35m_d2c_breif/config.py
2
+ output/sd35m_d2c_breif/checkpoints/epoch_40_step_121041.pth
sd35m_d2c_breif/metric_helper/epoch_45_step_136171.pth.txt ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ output/sd35m_d2c_breif/config.py
2
+ output/sd35m_d2c_breif/checkpoints/epoch_45_step_136171.pth
sd35m_d2c_breif/metric_helper/epoch_50_step_151301.pth.txt ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ output/sd35m_d2c_breif/config.py
2
+ output/sd35m_d2c_breif/checkpoints/epoch_50_step_151301.pth
sd35m_d2c_breif/metric_helper/epoch_55_step_166431.pth.txt ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ output/sd35m_d2c_breif/config.py
2
+ output/sd35m_d2c_breif/checkpoints/epoch_55_step_166431.pth
sd35m_d2c_breif/metric_helper/epoch_5_step_15131.pth.txt ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ output/sd35m_d2c_breif/config.py
2
+ output/sd35m_d2c_breif/checkpoints/epoch_5_step_15131.pth
sd35m_d2c_breif/metric_helper/epoch_60_step_181561.pth.txt ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ output/sd35m_d2c_breif/config.py
2
+ output/sd35m_d2c_breif/checkpoints/epoch_60_step_181561.pth
sd35m_d2c_breif/metric_helper/epoch_65_step_196691.pth.txt ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ output/sd35m_d2c_breif/config.py
2
+ output/sd35m_d2c_breif/checkpoints/epoch_65_step_196691.pth
sd35m_d2c_breif/metric_helper/epoch_70_step_211821.pth.txt ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ output/sd35m_d2c_breif/config.py
2
+ output/sd35m_d2c_breif/checkpoints/epoch_70_step_211821.pth
sd35m_d2c_breif/metric_helper/epoch_75_step_226951.pth.txt ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ output/sd35m_d2c_breif/config.py
2
+ output/sd35m_d2c_breif/checkpoints/epoch_75_step_226951.pth
sd35m_d2c_breif/metric_helper/epoch_80_step_242081.pth.txt ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ output/sd35m_d2c_breif/config.py
2
+ output/sd35m_d2c_breif/checkpoints/epoch_80_step_242081.pth
sd35m_d2c_breif/metric_helper/epoch_85_step_257211.pth.txt ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ output/sd35m_d2c_breif/config.py
2
+ output/sd35m_d2c_breif/checkpoints/epoch_85_step_257211.pth
sd35m_d2c_breif/metric_helper/epoch_90_step_272341.pth.txt ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ output/sd35m_d2c_breif/config.py
2
+ output/sd35m_d2c_breif/checkpoints/epoch_90_step_272341.pth
sd35m_d2c_breif/metric_helper/epoch_95_step_287471.pth.txt ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ output/sd35m_d2c_breif/config.py
2
+ output/sd35m_d2c_breif/checkpoints/epoch_95_step_287471.pth
sd35m_d2c_breif/train_log.log ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:153645fc5d6cca367237ed6ee2d07f3a711dc54b3a69f0c0c36f76b3d06e5667
3
+ size 76207919
sd35m_p2p_breif/checkpoints/epoch_100_step_302601.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:60ada01c0c01da3df533d9f7d8b37c86524e03ebfac518d563863b736a8789a4
3
+ size 18874194567
sd35m_p2p_breif/checkpoints/latest.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:60ada01c0c01da3df533d9f7d8b37c86524e03ebfac518d563863b736a8789a4
3
+ size 18874194567
sd35m_p2p_breif/config.yaml ADDED
@@ -0,0 +1,221 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ data:
2
+ data_dir:
3
+ - /home/work/shared-fi-datasets-01/users/hsiang.chen/Project/Datasets/IR
4
+ caption_proportion:
5
+ prompt: 1
6
+ external_caption_suffixes: []
7
+ external_clipscore_suffixes: []
8
+ clip_thr_temperature: 0.1
9
+ clip_thr: 25.0
10
+ del_img_clip_thr: 0.0
11
+ sort_dataset: false
12
+ load_text_feat: false
13
+ load_vae_feat: false
14
+ transform: default_train
15
+ type: IRImgDataset
16
+ image_size: 256
17
+ hq_only: false
18
+ valid_num: 0
19
+ data: null
20
+ extra: null
21
+ dset: train_brief
22
+ max_samples: null
23
+ model:
24
+ model: SD35M_P2P
25
+ model_pretrained: ./checkpoints/stable-diffusion-3.5-medium/sd3.5_medium.safetensors
26
+ shift: 3.0
27
+ teacher: null
28
+ input_channel: 16
29
+ image_size: 256
30
+ mixed_precision: bf16
31
+ fp32_attention: true
32
+ load_from: null
33
+ discriminator_model: null
34
+ teacher_model: null
35
+ teacher_model_weight_dtype: null
36
+ resume_from:
37
+ checkpoint: latest
38
+ load_ema: false
39
+ resume_optimizer: true
40
+ resume_lr_scheduler: true
41
+ aspect_ratio_type: ASPECT_RATIO_1024
42
+ multi_scale: false
43
+ pe_interpolation: 1.0
44
+ micro_condition: false
45
+ attn_type: linear
46
+ autocast_linear_attn: false
47
+ ffn_type: glumbconv
48
+ mlp_acts:
49
+ - silu
50
+ - silu
51
+ - null
52
+ mlp_ratio: 2.5
53
+ use_pe: false
54
+ pos_embed_type: sincos
55
+ qk_norm: false
56
+ class_dropout_prob: 0.1
57
+ linear_head_dim: 32
58
+ cross_norm: false
59
+ cross_attn_type: flash
60
+ logvar: false
61
+ cfg_scale: 4
62
+ cfg_embed: false
63
+ cfg_embed_scale: 1.0
64
+ guidance_type: classifier-free
65
+ pag_applied_layers:
66
+ - 8
67
+ ladd_multi_scale: true
68
+ head_block_ids: null
69
+ extra: null
70
+ vae:
71
+ vae_type: SDVAE
72
+ vae_pretrained: ./checkpoints/stable-diffusion-3.5-medium/sd3.5_medium.safetensors
73
+ weight_dtype: float32
74
+ scale_factor: 0.41407
75
+ vae_latent_dim: 16
76
+ vae_downsample_rate: 8
77
+ sample_posterior: true
78
+ extra: null
79
+ text_encoder:
80
+ text_encoder_name: sd35-text
81
+ text_encoder_pretrained: ./checkpoints/stable-diffusion-3.5-medium/text_encoders
82
+ caption_channels: 4096
83
+ y_norm: true
84
+ y_norm_scale_factor: 0.01
85
+ model_max_length: 300
86
+ chi_prompt:
87
+ - a photo of a cat
88
+ - Convenience store entrance at night. On the glass door, a vinyl decal reads
89
+ 'OPEN FOR QUALITY'. Inside, shelves and fluorescent lights; outside, a cyclist
90
+ passing by
91
+ - Sunrise beach, shallow tide washing over smooth sand. A piece of weathered driftwood
92
+ lies near the shoreline with a subtle branded text [SOS] on its surface; wet
93
+ sand reflections, micro-ripples, sun flare at horizon.
94
+ extra: null
95
+ scheduler:
96
+ train_sampling_steps: 1000
97
+ predict_flow_v: true
98
+ noise_schedule: linear_flow
99
+ pred_sigma: false
100
+ learn_sigma: true
101
+ vis_sampler: flow_dpm-solver
102
+ flow_shift: 3.0
103
+ weighting_scheme: logit_normal
104
+ weighting_scheme_discriminator: logit_normal_trigflow
105
+ add_noise_timesteps:
106
+ - 1.5708
107
+ logit_mean: 0.0
108
+ logit_std: 1.0
109
+ logit_mean_discriminator: 0.0
110
+ logit_std_discriminator: 1.0
111
+ sigma_data: 0.5
112
+ timestep_norm_scale_factor: 1.0
113
+ extra: null
114
+ train:
115
+ num_workers: 10
116
+ seed: 1229
117
+ train_batch_size: 4
118
+ num_epochs: 100
119
+ gradient_accumulation_steps: 8
120
+ grad_checkpointing: true
121
+ gradient_clip: 0.1
122
+ gc_step: 1
123
+ optimizer:
124
+ betas:
125
+ - 0.9
126
+ - 0.999
127
+ - 0.9999
128
+ eps:
129
+ - 1.0e-30
130
+ - 1.0e-16
131
+ lr: 5.0e-05
132
+ type: CAMEWrapper
133
+ weight_decay: 0.0
134
+ optimizer_D:
135
+ eps: 1.0e-10
136
+ lr: 0.0001
137
+ type: AdamW
138
+ weight_decay: 0.03
139
+ load_from_optimizer: false
140
+ load_from_lr_scheduler: false
141
+ resume_lr_scheduler: true
142
+ lr_schedule: cosine
143
+ lr_schedule_args:
144
+ num_warmup_steps: 2000
145
+ auto_lr:
146
+ rule: sqrt
147
+ eval_batch_size: 16
148
+ use_fsdp: false
149
+ use_flash_attn: false
150
+ eval_sampling_steps: 500
151
+ lora_rank: 4
152
+ log_interval: 1
153
+ mask_type: 'null'
154
+ mask_loss_coef: 0.0
155
+ load_mask_index: false
156
+ snr_loss: false
157
+ real_prompt_ratio: 1.0
158
+ early_stop_hours: 10000.0
159
+ save_image_epochs: 1
160
+ save_model_epochs: 5
161
+ save_model_steps: 500
162
+ visualize: true
163
+ null_embed_root: output/pretrained_models/
164
+ valid_prompt_embed_root: output/tmp_embed/
165
+ validation_prompts:
166
+ - dog
167
+ - portrait photo of a girl, photograph, highly detailed face, depth of field
168
+ - Self-portrait oil painting, a beautiful cyborg with golden hair, 8k
169
+ - Astronaut in a jungle, cold color palette, muted colors, detailed, 8k
170
+ - A photo of beautiful mountain with realistic sunset and blue lake, highly detailed,
171
+ masterpiece
172
+ local_save_vis: true
173
+ deterministic_validation: true
174
+ online_metric: false
175
+ eval_metric_step: 2000
176
+ online_metric_dir: metric_helper
177
+ work_dir: output/sd35m_p2p
178
+ skip_step: 0
179
+ loss_type: huber
180
+ huber_c: 0.001
181
+ num_ddim_timesteps: 50
182
+ ema_decay: 0.95
183
+ debug_nan: false
184
+ ema_update: false
185
+ ema_rate: 0.9999
186
+ tangent_warmup_steps: 10000
187
+ scm_cfg_scale:
188
+ - 1.0
189
+ cfg_interval: null
190
+ scm_logvar_loss: true
191
+ norm_invariant_to_spatial_dim: true
192
+ norm_same_as_512_scale: false
193
+ g_norm_constant: 0.1
194
+ g_norm_r: 1.0
195
+ show_gradient: false
196
+ lr_scale: null
197
+ adv_lambda: 1.0
198
+ scm_loss: true
199
+ scm_lambda: 1.0
200
+ loss_scale: 1.0
201
+ r1_penalty: false
202
+ r1_penalty_weight: 1.0e-05
203
+ diff_timesteps_D: true
204
+ suffix_checkpoints: disc
205
+ misaligned_pairs_D: false
206
+ discriminator_loss: cross entropy
207
+ largest_timestep: 1.5708
208
+ train_largest_timestep: false
209
+ largest_timestep_prob: 0.5
210
+ extra: null
211
+ controlnet: null
212
+ model_growth: null
213
+ work_dir: output/sd35m_p2p_breif
214
+ resume_from: latest
215
+ load_from: null
216
+ debug: true
217
+ caching: false
218
+ report_to: tensorboard
219
+ tracker_project_name: sana-baseline
220
+ name: tmp
221
+ loss_report_name: loss
sd35m_p2p_breif/metric_helper/epoch_100_step_302601.pth.txt ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ output/sd35m_p2p_breif/config.py
2
+ output/sd35m_p2p_breif/checkpoints/epoch_100_step_302601.pth
sd35m_p2p_breif/metric_helper/epoch_10_step_30261.pth.txt ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ output/sd35m_p2p_breif/config.py
2
+ output/sd35m_p2p_breif/checkpoints/epoch_10_step_30261.pth
sd35m_p2p_breif/metric_helper/epoch_15_step_45391.pth.txt ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ output/sd35m_p2p_breif/config.py
2
+ output/sd35m_p2p_breif/checkpoints/epoch_15_step_45391.pth
sd35m_p2p_breif/metric_helper/epoch_20_step_60521.pth.txt ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ output/sd35m_p2p_breif/config.py
2
+ output/sd35m_p2p_breif/checkpoints/epoch_20_step_60521.pth
sd35m_p2p_breif/metric_helper/epoch_25_step_75651.pth.txt ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ output/sd35m_p2p_breif/config.py
2
+ output/sd35m_p2p_breif/checkpoints/epoch_25_step_75651.pth
sd35m_p2p_breif/metric_helper/epoch_30_step_90781.pth.txt ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ output/sd35m_p2p_breif/config.py
2
+ output/sd35m_p2p_breif/checkpoints/epoch_30_step_90781.pth
sd35m_p2p_breif/metric_helper/epoch_35_step_105911.pth.txt ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ output/sd35m_p2p_breif/config.py
2
+ output/sd35m_p2p_breif/checkpoints/epoch_35_step_105911.pth
sd35m_p2p_breif/metric_helper/epoch_40_step_121041.pth.txt ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ output/sd35m_p2p_breif/config.py
2
+ output/sd35m_p2p_breif/checkpoints/epoch_40_step_121041.pth
sd35m_p2p_breif/metric_helper/epoch_45_step_136171.pth.txt ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ output/sd35m_p2p_breif/config.py
2
+ output/sd35m_p2p_breif/checkpoints/epoch_45_step_136171.pth
sd35m_p2p_breif/metric_helper/epoch_50_step_151301.pth.txt ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ output/sd35m_p2p_breif/config.py
2
+ output/sd35m_p2p_breif/checkpoints/epoch_50_step_151301.pth
sd35m_p2p_breif/metric_helper/epoch_55_step_166431.pth.txt ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ output/sd35m_p2p_breif/config.py
2
+ output/sd35m_p2p_breif/checkpoints/epoch_55_step_166431.pth
sd35m_p2p_breif/metric_helper/epoch_5_step_15131.pth.txt ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ output/sd35m_p2p_breif/config.py
2
+ output/sd35m_p2p_breif/checkpoints/epoch_5_step_15131.pth
sd35m_p2p_breif/metric_helper/epoch_60_step_181561.pth.txt ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ output/sd35m_p2p_breif/config.py
2
+ output/sd35m_p2p_breif/checkpoints/epoch_60_step_181561.pth
sd35m_p2p_breif/metric_helper/epoch_65_step_196691.pth.txt ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ output/sd35m_p2p_breif/config.py
2
+ output/sd35m_p2p_breif/checkpoints/epoch_65_step_196691.pth
sd35m_p2p_breif/metric_helper/epoch_70_step_211821.pth.txt ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ output/sd35m_p2p_breif/config.py
2
+ output/sd35m_p2p_breif/checkpoints/epoch_70_step_211821.pth
sd35m_p2p_breif/metric_helper/epoch_75_step_226951.pth.txt ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ output/sd35m_p2p_breif/config.py
2
+ output/sd35m_p2p_breif/checkpoints/epoch_75_step_226951.pth
sd35m_p2p_breif/metric_helper/epoch_80_step_242081.pth.txt ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ output/sd35m_p2p_breif/config.py
2
+ output/sd35m_p2p_breif/checkpoints/epoch_80_step_242081.pth
sd35m_p2p_breif/metric_helper/epoch_85_step_257211.pth.txt ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ output/sd35m_p2p_breif/config.py
2
+ output/sd35m_p2p_breif/checkpoints/epoch_85_step_257211.pth
sd35m_p2p_breif/metric_helper/epoch_90_step_272341.pth.txt ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ output/sd35m_p2p_breif/config.py
2
+ output/sd35m_p2p_breif/checkpoints/epoch_90_step_272341.pth
sd35m_p2p_breif/metric_helper/epoch_95_step_287471.pth.txt ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ output/sd35m_p2p_breif/config.py
2
+ output/sd35m_p2p_breif/checkpoints/epoch_95_step_287471.pth
sd35m_p2p_breif/train_log.log ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0ad6af5ef7235e94ea07e6f8e98be88fecdfa704aa8fb34710408c27809a4436
3
+ size 76200701