Finish training for 15k steps for phase 1 and 12.5k steps for phase 2
Browse files- phase1/FontDiffuser_training_phase_1_config.yaml +4 -4
- phase1/checkpoint_step_1375/content_encoder.safetensors +3 -0
- phase1/checkpoint_step_1375/style_encoder.safetensors +3 -0
- phase1/checkpoint_step_1375/training_state.pt +3 -0
- phase1/checkpoint_step_1375/unet.safetensors +3 -0
- phase1/checkpoint_step_2750/content_encoder.safetensors +3 -0
- phase1/checkpoint_step_2750/style_encoder.safetensors +3 -0
- phase1/checkpoint_step_2750/training_state.pt +3 -0
- phase1/checkpoint_step_2750/unet.safetensors +3 -0
- phase1/final/content_encoder.safetensors +3 -0
- phase1/final/style_encoder.safetensors +3 -0
- phase1/final/training_state.pt +3 -0
- phase1/final/unet.safetensors +3 -0
- phase2/FontDiffuser_training_phase_2_config.yaml +6 -6
- phase2/checkpoint_step_1250/content_encoder.safetensors +3 -0
- phase2/checkpoint_step_1250/scr.safetensors +3 -0
- phase2/checkpoint_step_1250/style_encoder.safetensors +3 -0
- phase2/checkpoint_step_1250/training_state.pt +3 -0
- phase2/checkpoint_step_1250/unet.safetensors +3 -0
- phase2/checkpoint_step_2500/content_encoder.safetensors +3 -0
- phase2/checkpoint_step_2500/scr.safetensors +3 -0
- phase2/checkpoint_step_2500/style_encoder.safetensors +3 -0
- phase2/checkpoint_step_2500/training_state.pt +3 -0
- phase2/checkpoint_step_2500/unet.safetensors +3 -0
phase1/FontDiffuser_training_phase_1_config.yaml
CHANGED
|
@@ -5,7 +5,7 @@ adam_weight_decay: 0.01
|
|
| 5 |
algorithm_type: dpmsolver++
|
| 6 |
beta_scheduler: scaled_linear
|
| 7 |
channel_attn: true
|
| 8 |
-
ckpt_interval:
|
| 9 |
content_encoder_downsample_size: 3
|
| 10 |
content_image_size: !!python/tuple
|
| 11 |
- 96
|
|
@@ -28,7 +28,7 @@ logging_dir: logs
|
|
| 28 |
lr_scheduler: cosine
|
| 29 |
lr_warmup_steps: 200
|
| 30 |
max_grad_norm: 1.0
|
| 31 |
-
max_train_steps:
|
| 32 |
method: multistep
|
| 33 |
mixed_precision: fp16
|
| 34 |
mode: refinement
|
|
@@ -42,7 +42,7 @@ order: 2
|
|
| 42 |
output_dir: outputs/FontDiffuser/phase1
|
| 43 |
perceptual_coefficient: 0.05
|
| 44 |
phase_1: true
|
| 45 |
-
phase_1_ckpt_dir: ckpt/phase1/
|
| 46 |
phase_2: false
|
| 47 |
report_to: wandb
|
| 48 |
resolution: 96
|
|
@@ -61,7 +61,7 @@ style_start_channel: 64
|
|
| 61 |
t_end: null
|
| 62 |
t_start: null
|
| 63 |
temperature: 0.07
|
| 64 |
-
train_batch_size:
|
| 65 |
unet_channels: !!python/tuple
|
| 66 |
- 64
|
| 67 |
- 128
|
|
|
|
| 5 |
algorithm_type: dpmsolver++
|
| 6 |
beta_scheduler: scaled_linear
|
| 7 |
channel_attn: true
|
| 8 |
+
ckpt_interval: 1375
|
| 9 |
content_encoder_downsample_size: 3
|
| 10 |
content_image_size: !!python/tuple
|
| 11 |
- 96
|
|
|
|
| 28 |
lr_scheduler: cosine
|
| 29 |
lr_warmup_steps: 200
|
| 30 |
max_grad_norm: 1.0
|
| 31 |
+
max_train_steps: 2750
|
| 32 |
method: multistep
|
| 33 |
mixed_precision: fp16
|
| 34 |
mode: refinement
|
|
|
|
| 42 |
output_dir: outputs/FontDiffuser/phase1
|
| 43 |
perceptual_coefficient: 0.05
|
| 44 |
phase_1: true
|
| 45 |
+
phase_1_ckpt_dir: ckpt/phase1/checkpoint_step_2250
|
| 46 |
phase_2: false
|
| 47 |
report_to: wandb
|
| 48 |
resolution: 96
|
|
|
|
| 61 |
t_end: null
|
| 62 |
t_start: null
|
| 63 |
temperature: 0.07
|
| 64 |
+
train_batch_size: 8
|
| 65 |
unet_channels: !!python/tuple
|
| 66 |
- 64
|
| 67 |
- 128
|
phase1/checkpoint_step_1375/content_encoder.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9806be40c09d20fda5d607f8ba3f5f689f805dbd2e788d11c87a3df39c0781d4
|
| 3 |
+
size 4756580
|
phase1/checkpoint_step_1375/style_encoder.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:01a8033abe1120e379d7a41c9fc1edbdf125a4e37f6808d0cb0c03c0181fa9cf
|
| 3 |
+
size 82394556
|
phase1/checkpoint_step_1375/training_state.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1b519d7b8921a7af8c8247c86ef48954d7b0b4c63a8c7e22db4e833801a7c7a2
|
| 3 |
+
size 1206964193
|
phase1/checkpoint_step_1375/unet.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e91d3ab1f8058b7a08992dc0d8016c26032808ccee24e9c95fec0825f4e23110
|
| 3 |
+
size 314927748
|
phase1/checkpoint_step_2750/content_encoder.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a366571f4f39c4541f5f9bc67432c37772ef1d768c856caaad788d2849da8223
|
| 3 |
+
size 4756580
|
phase1/checkpoint_step_2750/style_encoder.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8e930895865ba9e262970c41f5f6d9d0cbb2a9bb519d151256fb467520f25738
|
| 3 |
+
size 82394556
|
phase1/checkpoint_step_2750/training_state.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3b166b583df7566226e28bb235ce9252a501177256d7d6283eef019f2d6b03f4
|
| 3 |
+
size 1206964193
|
phase1/checkpoint_step_2750/unet.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2319291039dbb4c8b66def131846bb2bb26c11613430782450000c1b7173fabf
|
| 3 |
+
size 314927748
|
phase1/final/content_encoder.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a366571f4f39c4541f5f9bc67432c37772ef1d768c856caaad788d2849da8223
|
| 3 |
+
size 4756580
|
phase1/final/style_encoder.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8e930895865ba9e262970c41f5f6d9d0cbb2a9bb519d151256fb467520f25738
|
| 3 |
+
size 82394556
|
phase1/final/training_state.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3b166b583df7566226e28bb235ce9252a501177256d7d6283eef019f2d6b03f4
|
| 3 |
+
size 1206964193
|
phase1/final/unet.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2319291039dbb4c8b66def131846bb2bb26c11613430782450000c1b7173fabf
|
| 3 |
+
size 314927748
|
phase2/FontDiffuser_training_phase_2_config.yaml
CHANGED
|
@@ -5,7 +5,7 @@ adam_weight_decay: 0.01
|
|
| 5 |
algorithm_type: dpmsolver++
|
| 6 |
beta_scheduler: scaled_linear
|
| 7 |
channel_attn: true
|
| 8 |
-
ckpt_interval:
|
| 9 |
content_encoder_downsample_size: 3
|
| 10 |
content_image_size: !!python/tuple
|
| 11 |
- 96
|
|
@@ -28,7 +28,7 @@ logging_dir: logs
|
|
| 28 |
lr_scheduler: cosine
|
| 29 |
lr_warmup_steps: 100
|
| 30 |
max_grad_norm: 1.0
|
| 31 |
-
max_train_steps:
|
| 32 |
method: multistep
|
| 33 |
mixed_precision: 'no'
|
| 34 |
mode: refinement
|
|
@@ -40,17 +40,17 @@ num_workers: 4
|
|
| 40 |
offset_coefficient: 0.6
|
| 41 |
order: 2
|
| 42 |
output_dir: outputs/FontDiffuser/phase2
|
| 43 |
-
perceptual_coefficient: 0.
|
| 44 |
phase_1: false
|
| 45 |
-
phase_1_ckpt_dir:
|
| 46 |
phase_2: true
|
| 47 |
report_to: wandb
|
| 48 |
resolution: 96
|
| 49 |
resume_from_checkpoint: null
|
| 50 |
save_full_model: false
|
| 51 |
-
sc_coefficient: 0.
|
| 52 |
scale_lr: false
|
| 53 |
-
scr_ckpt_path: ckpt/
|
| 54 |
scr_image_size: 96
|
| 55 |
seed: 123
|
| 56 |
skip_type: time_uniform
|
|
|
|
| 5 |
algorithm_type: dpmsolver++
|
| 6 |
beta_scheduler: scaled_linear
|
| 7 |
channel_attn: true
|
| 8 |
+
ckpt_interval: 1250
|
| 9 |
content_encoder_downsample_size: 3
|
| 10 |
content_image_size: !!python/tuple
|
| 11 |
- 96
|
|
|
|
| 28 |
lr_scheduler: cosine
|
| 29 |
lr_warmup_steps: 100
|
| 30 |
max_grad_norm: 1.0
|
| 31 |
+
max_train_steps: 2500
|
| 32 |
method: multistep
|
| 33 |
mixed_precision: 'no'
|
| 34 |
mode: refinement
|
|
|
|
| 40 |
offset_coefficient: 0.6
|
| 41 |
order: 2
|
| 42 |
output_dir: outputs/FontDiffuser/phase2
|
| 43 |
+
perceptual_coefficient: 0.08
|
| 44 |
phase_1: false
|
| 45 |
+
phase_1_ckpt_dir: outputs/FontDiffuser/phase1/final
|
| 46 |
phase_2: true
|
| 47 |
report_to: wandb
|
| 48 |
resolution: 96
|
| 49 |
resume_from_checkpoint: null
|
| 50 |
save_full_model: false
|
| 51 |
+
sc_coefficient: 0.1
|
| 52 |
scale_lr: false
|
| 53 |
+
scr_ckpt_path: ckpt/phase1/checkpoint_step_2250/scr.safetensors
|
| 54 |
scr_image_size: 96
|
| 55 |
seed: 123
|
| 56 |
skip_type: time_uniform
|
phase2/checkpoint_step_1250/content_encoder.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9e32fb85fe179b776abc95e1c534111f08456fd9cb14fb304228c0b3dd56adc5
|
| 3 |
+
size 4756580
|
phase2/checkpoint_step_1250/scr.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2e7eff5bad07d6f39d2deb4a19fdae042cfc311d3a0dde2d74bb718feb79727a
|
| 3 |
+
size 284227968
|
phase2/checkpoint_step_1250/style_encoder.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:cfd00b48d9f8dd0fea9eab31c34bc855d3325c2ab80ee6e93b387a4c4567eaf1
|
| 3 |
+
size 82394556
|
phase2/checkpoint_step_1250/training_state.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:61e37db8f92c07586ca55ae7f45af97cedc5b40f434b46e2517a195512021561
|
| 3 |
+
size 1206964193
|
phase2/checkpoint_step_1250/unet.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0a81b7b9d177e669b7dd57195e6c0e0e7ab2d50764aafa88f40aeba7512ccd5a
|
| 3 |
+
size 314927748
|
phase2/checkpoint_step_2500/content_encoder.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7a41c89951882f95ff82bd264dd7310d57ac69087d560797271feccda5cb8d1b
|
| 3 |
+
size 4756580
|
phase2/checkpoint_step_2500/scr.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1737df7040325d580d21a3bb8ff00c2be487757730cc93adb2a061e4dac7f376
|
| 3 |
+
size 284227968
|
phase2/checkpoint_step_2500/style_encoder.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a1aba03215480affb3b2d2c00dffbcf708522a5220fd98ede864898318fbc809
|
| 3 |
+
size 82394556
|
phase2/checkpoint_step_2500/training_state.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4ea9b27282dbcd8b555f7c1234b449b50f2d423f18d2a914af7d772d19032341
|
| 3 |
+
size 746024960
|
phase2/checkpoint_step_2500/unet.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6fa345d315f0759fad7367aa9de74fd7a30d8be935cd7393692c0b53b4b104e9
|
| 3 |
+
size 314927748
|