Upload folder using huggingface_hub

pre-trained/autoencoder/svg_autoencoder-P-stage1.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1083c1e2ee4cabc24c0a7a51eb98a3a8177589658b01fe691048c3532f18680f
+size 357108579

pre-trained/autoencoder/svg_autoencoder-P-stage2.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:457124c0a92c1c4b976b7e2a6f67cb8bba19e83ac14ae1f70de502337535b117
+size 357108579

pre-trained/autoencoder/svg_autoencoder-P-stage3.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:23f8f42e2384488563a90ad23691e17bee8e33d18bb6ce560a7ed108a5424285
+size 357108579

pre-trained/autoencoder/svg_autoencoder-R-stage1.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:60cc3612d173d57338612f30edb71043f4a009aeb6f8ea9f5ecbd2a0d1de0329
+size 445917171

pre-trained/autoencoder/svg_autoencoder-R-stage2.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ed64535c7809db4745c979fda99691b141b37a3a203ca2dffd325f343130a964
+size 445917171

pre-trained/autoencoder/svg_autoencoder-R-stage3.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f232b8d5feac07287e225b1d1b6a2ca54226858409834a838cf397df52b592df
+size 445917171

pre-trained/autoencoder/svg_autoencoder_P_stage1_256.yaml

@@ -0,0 +1,52 @@
+training:
+  logdir: "logs"
+  scale_lr: false
+  accumulate_grad_batches: 1
+  save_every_n_train_steps: 20000
+  save_every_n_epochs: 1
+
+  trainer:
+    devices: 8
+    num_nodes: 1
+    strategy: "ddp_find_unused_parameters_true"
+    accelerator: "gpu"
+    max_epochs: 1000
+    precision: 32
+
+# Model configuration
+model:
+  init_weight: pre-trained/autoencoder/svg_autoencoder-P-stage1.ckpt
+  base_learning_rate: 1.0e-04
+  target: ldm.models.dinov3_decoder_native_resolution.DinoDecoder
+  params:    
+    ckpt_path: pre-trained/autoencoder/svg_autoencoder-P-stage1.ckpt
+    embed_dim: 32
+    lossconfig:
+      target: "ldm.modules.losses.LPIPSWithDiscriminatorDecoder"
+      params:
+        disc_start: 5000
+        disc_weight: 0.5
+    ddconfig:
+      double_z: true
+      z_channels: 384
+      resolution: 256
+      in_channels: 3
+      out_ch: 3
+      ch: 128
+      ch_mult:
+      - 1
+      - 1
+      - 2
+      - 2
+      - 4
+      num_res_blocks: 2
+      attn_resolutions: []
+      dropout: 0.0
+
+    dinoconfig:
+      dinov3_location: dinov3
+      model_name: dinov3_vits16plus
+      weights: pre-trained/dinov3_vits16plus_pretrain_lvd1689m-4057cbaa.pth
+    extra_vit_config: null
+
+

pre-trained/autoencoder/svg_autoencoder_P_stage2_512.yaml

@@ -0,0 +1,52 @@
+training:
+  logdir: "logs"
+  scale_lr: false
+  accumulate_grad_batches: 1
+  save_every_n_train_steps: 20000
+  save_every_n_epochs: 1
+
+  trainer:
+    devices: 8
+    num_nodes: 1
+    strategy: "ddp_find_unused_parameters_true"
+    accelerator: "gpu"
+    max_epochs: 1000
+    precision: 32
+
+# Model configuration
+model:
+  init_weight: pre-trained/autoencoder/svg_autoencoder-P-stage2.ckpt
+  base_learning_rate: 1.0e-04
+  target: ldm.models.dinov3_decoder_native_resolution.DinoDecoder
+  params:    
+    ckpt_path: pre-trained/autoencoder/svg_autoencoder-P-stage2.ckpt
+    embed_dim: 32
+    lossconfig:
+      target: "ldm.modules.losses.LPIPSWithDiscriminatorDecoder"
+      params:
+        disc_start: 5000
+        disc_weight: 0.5
+    ddconfig:
+      double_z: true
+      z_channels: 384
+      resolution: 256
+      in_channels: 3
+      out_ch: 3
+      ch: 128
+      ch_mult:
+      - 1
+      - 1
+      - 2
+      - 2
+      - 4
+      num_res_blocks: 2
+      attn_resolutions: []
+      dropout: 0.0
+
+    dinoconfig:
+      dinov3_location: dinov3
+      model_name: dinov3_vits16plus
+      weights: pre-trained/dinov3_vits16plus_pretrain_lvd1689m-4057cbaa.pth
+    extra_vit_config: null
+
+

pre-trained/autoencoder/svg_autoencoder_P_stage3_1024.yaml

@@ -0,0 +1,52 @@
+training:
+  logdir: "logs"
+  scale_lr: false
+  accumulate_grad_batches: 1
+  save_every_n_train_steps: 20000
+  save_every_n_epochs: 1
+
+  trainer:
+    devices: 8
+    num_nodes: 1
+    strategy: "ddp_find_unused_parameters_true"
+    accelerator: "gpu"
+    max_epochs: 1000
+    precision: 32
+
+# Model configuration
+model:
+  init_weight: pre-trained/autoencoder/svg_autoencoder-P-stage3.ckpt
+  base_learning_rate: 1.0e-04
+  target: ldm.models.dinov3_decoder_native_resolution.DinoDecoder
+  params:    
+    ckpt_path: pre-trained/autoencoder/svg_autoencoder-P-stage3.ckpt
+    embed_dim: 32
+    lossconfig:
+      target: "ldm.modules.losses.LPIPSWithDiscriminatorDecoder"
+      params:
+        disc_start: 5000
+        disc_weight: 0.5
+    ddconfig:
+      double_z: true
+      z_channels: 384
+      resolution: 256
+      in_channels: 3
+      out_ch: 3
+      ch: 128
+      ch_mult:
+      - 1
+      - 1
+      - 2
+      - 2
+      - 4
+      num_res_blocks: 2
+      attn_resolutions: []
+      dropout: 0.0
+
+    dinoconfig:
+      dinov3_location: dinov3
+      model_name: dinov3_vits16plus
+      weights: pre-trained/dinov3_vits16plus_pretrain_lvd1689m-4057cbaa.pth
+    extra_vit_config: null
+
+

pre-trained/autoencoder/svg_autoencoder_R_stage1_256.yaml

@@ -0,0 +1,52 @@
+training:
+  logdir: "logs"
+  scale_lr: false
+  accumulate_grad_batches: 1
+  save_every_n_train_steps: 20000
+  save_every_n_epochs: 1
+
+  trainer:
+    devices: 8
+    num_nodes: 1
+    strategy: "ddp_find_unused_parameters_true"
+    accelerator: "gpu"
+    max_epochs: 1000
+    precision: 32
+
+# Model configuration
+model:
+  init_weight: pre-trained/autoencoder/svg_autoencoder-R-stage1.ckpt
+  base_learning_rate: 1.0e-04
+  target: ldm.models.dinov3_decoder_native_resolution.DinoDecoder
+  params:    
+    ckpt_path: pre-trained/autoencoder/svg_autoencoder-R-stage1.ckpt
+    embed_dim: 32
+    lossconfig:
+      target: "ldm.modules.losses.LPIPSWithDiscriminatorDecoder"
+      params:
+        disc_start: 5000
+        disc_weight: 0.5
+    ddconfig:
+      double_z: true
+      z_channels: 384
+      resolution: 256
+      in_channels: 3
+      out_ch: 3
+      ch: 128
+      ch_mult:
+      - 1
+      - 1
+      - 2
+      - 2
+      - 4
+      num_res_blocks: 2
+      attn_resolutions: []
+      dropout: 0.0
+
+    dinoconfig:
+      dinov3_location: dinov3
+      model_name: dinov3_vits16plus
+      weights: pre-trained/dinov3_vits16plus_pretrain_lvd1689m-4057cbaa.pth
+    extra_vit_config: null
+
+

pre-trained/autoencoder/svg_autoencoder_R_stage2_512.yaml

@@ -0,0 +1,52 @@
+training:
+  logdir: "logs"
+  scale_lr: false
+  accumulate_grad_batches: 1
+  save_every_n_train_steps: 20000
+  save_every_n_epochs: 1
+
+  trainer:
+    devices: 8
+    num_nodes: 1
+    strategy: "ddp_find_unused_parameters_true"
+    accelerator: "gpu"
+    max_epochs: 1000
+    precision: 32
+
+# Model configuration
+model:
+  init_weight: pre-trained/autoencoder/svg_autoencoder-R-stage2.ckpt
+  base_learning_rate: 1.0e-04
+  target: ldm.models.dinov3_decoder_native_resolution.DinoDecoder
+  params:    
+    ckpt_path: pre-trained/autoencoder/svg_autoencoder-R-stage2.ckpt
+    embed_dim: 32
+    lossconfig:
+      target: "ldm.modules.losses.LPIPSWithDiscriminatorDecoder"
+      params:
+        disc_start: 5000
+        disc_weight: 0.5
+    ddconfig:
+      double_z: true
+      z_channels: 384
+      resolution: 256
+      in_channels: 3
+      out_ch: 3
+      ch: 128
+      ch_mult:
+      - 1
+      - 1
+      - 2
+      - 2
+      - 4
+      num_res_blocks: 2
+      attn_resolutions: []
+      dropout: 0.0
+
+    dinoconfig:
+      dinov3_location: dinov3
+      model_name: dinov3_vits16plus
+      weights: pre-trained/dinov3_vits16plus_pretrain_lvd1689m-4057cbaa.pth
+    extra_vit_config: null
+
+

pre-trained/autoencoder/svg_autoencoder_R_stage3_1024.yaml

@@ -0,0 +1,52 @@
+training:
+  logdir: "logs"
+  scale_lr: false
+  accumulate_grad_batches: 1
+  save_every_n_train_steps: 20000
+  save_every_n_epochs: 1
+
+  trainer:
+    devices: 8
+    num_nodes: 1
+    strategy: "ddp_find_unused_parameters_true"
+    accelerator: "gpu"
+    max_epochs: 1000
+    precision: 32
+
+# Model configuration
+model:
+  init_weight: pre-trained/autoencoder/svg_autoencoder-R-stage3.ckpt
+  base_learning_rate: 1.0e-04
+  target: ldm.models.dinov3_decoder_native_resolution.DinoDecoder
+  params:    
+    ckpt_path: pre-trained/autoencoder/svg_autoencoder-R-stage3.ckpt
+    embed_dim: 32
+    lossconfig:
+      target: "ldm.modules.losses.LPIPSWithDiscriminatorDecoder"
+      params:
+        disc_start: 5000
+        disc_weight: 0.5
+    ddconfig:
+      double_z: true
+      z_channels: 384
+      resolution: 256
+      in_channels: 3
+      out_ch: 3
+      ch: 128
+      ch_mult:
+      - 1
+      - 1
+      - 2
+      - 2
+      - 4
+      num_res_blocks: 2
+      attn_resolutions: []
+      dropout: 0.0
+
+    dinoconfig:
+      dinov3_location: dinov3
+      model_name: dinov3_vits16plus
+      weights: pre-trained/dinov3_vits16plus_pretrain_lvd1689m-4057cbaa.pth
+    extra_vit_config: null
+
+