Add checkpoints for celeba(32/64/256)

celeba256/celeba256_eps1.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:074ee91513ae2c57da811ef8ffd5661ee3b6679edc704877563e14225bc9842a
+size 3182184173

celeba256/celeba256_eps1.yaml

@@ -0,0 +1,111 @@
+model:
+  base_learning_rate: 1.0e-07
+  target: ldm.models.diffusion.ddpm.LatentDiffusion
+  params:
+    ckpt_path: models/ldm/cin256-v2/model.ckpt
+    ignore_keys:
+    - cond_stage_model
+    linear_start: 0.0015
+    linear_end: 0.0195
+    num_timesteps_cond: 1
+    log_every_t: 200
+    timesteps: 1000
+    first_stage_key: image
+    cond_stage_key: class_label
+    image_size: 64
+    channels: 3
+    train_condition_only: true
+    conditioning_key: crossattn
+    monitor: val/loss
+    use_ema: false
+    dp_config:
+      enabled: true
+      epsilon: 1
+      delta: 1.0e-05
+      max_grad_norm: 0.0005
+      poisson_sampling: true
+      max_batch_size: 15
+    unet_config:
+      target: ldm.modules.diffusionmodules.openaimodel.UNetModel
+      params:
+        image_size: 64
+        in_channels: 3
+        out_channels: 3
+        model_channels: 192
+        attention_resolutions:
+        - 8
+        - 4
+        - 2
+        num_res_blocks: 2
+        channel_mult:
+        - 1
+        - 2
+        - 3
+        - 5
+        num_heads: 1
+        use_spatial_transformer: true
+        transformer_depth: 1
+        context_dim: 512
+    first_stage_config:
+      target: ldm.models.autoencoder.VQModelInterface
+      params:
+        embed_dim: 3
+        n_embed: 8192
+        ddconfig:
+          double_z: false
+          z_channels: 3
+          resolution: 256
+          in_channels: 3
+          out_ch: 3
+          ch: 128
+          ch_mult:
+          - 1
+          - 2
+          - 4
+          num_res_blocks: 2
+          attn_resolutions: []
+          dropout: 0.0
+        lossconfig:
+          target: torch.nn.Identity
+    cond_stage_config:
+      target: ldm.modules.encoders.modules.ClassEmbedder
+      params:
+        n_classes: 2
+        embed_dim: 512
+        key: class_label
+    ablation_blocks: -1
+data:
+  target: main.DataModuleFromConfig
+  params:
+    batch_size: 2048
+    num_workers: 6
+    wrap: false
+    train:
+      target: ldm.data.celebahq.CelebAHQ
+      params:
+        split: train
+        target_type: class
+        class_attrs:
+        - Male
+        size: 256
+    validation:
+      target: ldm.data.celebahq.CelebAHQ
+      params:
+        split: test
+        target_type: class
+        class_attrs:
+        - Male
+        size: 256
+lightning:
+  callbacks:
+    image_logger:
+      target: main.ImageLogger
+      params:
+        batch_frequency: 25000
+        max_images: 8
+        increase_log_steps: false
+  trainer:
+    benchmark: true
+    max_epochs: 50
+    accelerator: gpu
+    gpus: 0,

celeba256/celeba256_eps10.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:78ecd018a1856f019bec1b306ccddc59a3c4a231570e8639c259abd9062e3a59
+size 3182184173

celeba256/celeba256_eps10.yaml

@@ -0,0 +1,111 @@
+model:
+  base_learning_rate: 1.0e-07
+  target: ldm.models.diffusion.ddpm.LatentDiffusion
+  params:
+    ckpt_path: models/ldm/cin256-v2/model.ckpt
+    ignore_keys:
+    - cond_stage_model
+    linear_start: 0.0015
+    linear_end: 0.0195
+    num_timesteps_cond: 1
+    log_every_t: 200
+    timesteps: 1000
+    first_stage_key: image
+    cond_stage_key: class_label
+    image_size: 64
+    channels: 3
+    train_condition_only: true
+    conditioning_key: crossattn
+    monitor: val/loss
+    use_ema: false
+    dp_config:
+      enabled: true
+      epsilon: 10
+      delta: 1.0e-05
+      max_grad_norm: 0.0005
+      poisson_sampling: true
+      max_batch_size: 15
+    unet_config:
+      target: ldm.modules.diffusionmodules.openaimodel.UNetModel
+      params:
+        image_size: 64
+        in_channels: 3
+        out_channels: 3
+        model_channels: 192
+        attention_resolutions:
+        - 8
+        - 4
+        - 2
+        num_res_blocks: 2
+        channel_mult:
+        - 1
+        - 2
+        - 3
+        - 5
+        num_heads: 1
+        use_spatial_transformer: true
+        transformer_depth: 1
+        context_dim: 512
+    first_stage_config:
+      target: ldm.models.autoencoder.VQModelInterface
+      params:
+        embed_dim: 3
+        n_embed: 8192
+        ddconfig:
+          double_z: false
+          z_channels: 3
+          resolution: 256
+          in_channels: 3
+          out_ch: 3
+          ch: 128
+          ch_mult:
+          - 1
+          - 2
+          - 4
+          num_res_blocks: 2
+          attn_resolutions: []
+          dropout: 0.0
+        lossconfig:
+          target: torch.nn.Identity
+    cond_stage_config:
+      target: ldm.modules.encoders.modules.ClassEmbedder
+      params:
+        n_classes: 2
+        embed_dim: 512
+        key: class_label
+    ablation_blocks: -1
+data:
+  target: main.DataModuleFromConfig
+  params:
+    batch_size: 2048
+    num_workers: 6
+    wrap: false
+    train:
+      target: ldm.data.celebahq.CelebAHQ
+      params:
+        split: train
+        target_type: class
+        class_attrs:
+        - Male
+        size: 256
+    validation:
+      target: ldm.data.celebahq.CelebAHQ
+      params:
+        split: test
+        target_type: class
+        class_attrs:
+        - Male
+        size: 256
+lightning:
+  callbacks:
+    image_logger:
+      target: main.ImageLogger
+      params:
+        batch_frequency: 25000
+        max_images: 8
+        increase_log_steps: false
+  trainer:
+    benchmark: true
+    max_epochs: 50
+    accelerator: gpu
+    gpus: 0,

celeba256/celeba256_eps5.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ea57477a701530fb09507b5295e3c291bba91678b3e45088a09ef8ca2accb09f
+size 3182184173

celeba256/celeba256_eps5.yaml

@@ -0,0 +1,111 @@
+model:
+  base_learning_rate: 1.0e-07
+  target: ldm.models.diffusion.ddpm.LatentDiffusion
+  params:
+    ckpt_path: models/ldm/cin256-v2/model.ckpt
+    ignore_keys:
+    - cond_stage_model
+    linear_start: 0.0015
+    linear_end: 0.0195
+    num_timesteps_cond: 1
+    log_every_t: 200
+    timesteps: 1000
+    first_stage_key: image
+    cond_stage_key: class_label
+    image_size: 64
+    channels: 3
+    train_condition_only: true
+    conditioning_key: crossattn
+    monitor: val/loss
+    use_ema: false
+    dp_config:
+      enabled: true
+      epsilon: 5
+      delta: 1.0e-05
+      max_grad_norm: 0.0005
+      poisson_sampling: true
+      max_batch_size: 15
+    unet_config:
+      target: ldm.modules.diffusionmodules.openaimodel.UNetModel
+      params:
+        image_size: 64
+        in_channels: 3
+        out_channels: 3
+        model_channels: 192
+        attention_resolutions:
+        - 8
+        - 4
+        - 2
+        num_res_blocks: 2
+        channel_mult:
+        - 1
+        - 2
+        - 3
+        - 5
+        num_heads: 1
+        use_spatial_transformer: true
+        transformer_depth: 1
+        context_dim: 512
+    first_stage_config:
+      target: ldm.models.autoencoder.VQModelInterface
+      params:
+        embed_dim: 3
+        n_embed: 8192
+        ddconfig:
+          double_z: false
+          z_channels: 3
+          resolution: 256
+          in_channels: 3
+          out_ch: 3
+          ch: 128
+          ch_mult:
+          - 1
+          - 2
+          - 4
+          num_res_blocks: 2
+          attn_resolutions: []
+          dropout: 0.0
+        lossconfig:
+          target: torch.nn.Identity
+    cond_stage_config:
+      target: ldm.modules.encoders.modules.ClassEmbedder
+      params:
+        n_classes: 2
+        embed_dim: 512
+        key: class_label
+    ablation_blocks: -1
+data:
+  target: main.DataModuleFromConfig
+  params:
+    batch_size: 2048
+    num_workers: 6
+    wrap: false
+    train:
+      target: ldm.data.celebahq.CelebAHQ
+      params:
+        split: train
+        target_type: class
+        class_attrs:
+        - Male
+        size: 256
+    validation:
+      target: ldm.data.celebahq.CelebAHQ
+      params:
+        split: test
+        target_type: class
+        class_attrs:
+        - Male
+        size: 256
+lightning:
+  callbacks:
+    image_logger:
+      target: main.ImageLogger
+      params:
+        batch_frequency: 25000
+        max_images: 8
+        increase_log_steps: false
+  trainer:
+    benchmark: true
+    max_epochs: 50
+    accelerator: gpu
+    gpus: 0,

celeba32/celeba32_eps1.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:158e5b0ca17e4a4ded118643f44bcc46b0830885b900aafb75ca0bad719876d4
+size 1570608894

celeba32/celeba32_eps1.yaml

@@ -0,0 +1,104 @@
+model:
+  base_learning_rate: 5.0e-07
+  target: ldm.models.diffusion.ddpm.LatentDiffusion
+  params:
+    ckpt_path: ...
+    ignore_keys:
+    - cond_stage_model
+    linear_start: 0.0015
+    linear_end: 0.0195
+    timesteps: 1000
+    num_timesteps_cond: 1
+    first_stage_key: image
+    image_size: 16
+    channels: 3
+    cond_stage_trainable: false
+    concat_mode: false
+    scale_by_std: true
+    train_attention_only: true
+    attention_flag: self
+    monitor: val/loss_simple_ema
+    log_every_t: 200
+    dp_config:
+      enabled: true
+      epsilon: 1
+      delta: 1.0e-06
+      max_grad_norm: 0.0005
+      poisson_sampling: true
+      max_batch_size: 100
+    unet_config:
+      target: ldm.modules.diffusionmodules.openaimodel.UNetModel
+      params:
+        image_size: 16
+        in_channels: 3
+        out_channels: 3
+        model_channels: 192
+        attention_resolutions:
+        - 4
+        - 2
+        - 1
+        num_res_blocks: 2
+        channel_mult:
+        - 1
+        - 2
+        - 4
+        num_head_channels: 32
+        use_scale_shift_norm: true
+        resblock_updown: true
+    first_stage_config:
+      target: ldm.models.autoencoder.AutoencoderKL
+      params:
+        monitor: val/rec_loss
+        embed_dim: 3
+        lossconfig:
+          target: ldm.modules.losses.LPIPSWithDiscriminator
+          params:
+            disc_start: 50001
+            kl_weight: 1.0e-06
+            disc_weight: 0.5
+        ddconfig:
+          double_z: true
+          z_channels: 3
+          resolution: 32
+          in_channels: 3
+          out_ch: 3
+          ch: 128
+          ch_mult:
+          - 1
+          - 2
+          num_res_blocks: 2
+          attn_resolutions:
+          - 16
+          - 8
+          dropout: 0.0
+    cond_stage_config: __is_unconditional__
+    ablation_blocks: -1
+data:
+  target: main.DataModuleFromConfig
+  params:
+    batch_size: 2048
+    num_workers: 4
+    wrap: false
+    train:
+      target: ldm.data.celeba.CelebATrain
+      params:
+        config:
+          size: 32
+    validation:
+      target: ldm.data.celeba.CelebAValidation
+      params:
+        config:
+          size: 32
+lightning:
+  callbacks:
+    image_logger:
+      target: main.ImageLogger
+      params:
+        batch_frequency: 500
+        max_images: 8
+        increase_log_steps: false
+  trainer:
+    benchmark: true
+    max_epochs: 20
+    accelerator: gpu
+    gpus: 0,

celeba32/celeba32_eps10.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c4b0db5eb9e4a79f0e2fe4025e3714a93c2b7dfbcefc247bfd3992600113df33
+size 1570608894

celeba32/celeba32_eps10.yaml

@@ -0,0 +1,104 @@
+model:
+  base_learning_rate: 5.0e-07
+  target: ldm.models.diffusion.ddpm.LatentDiffusion
+  params:
+    ckpt_path: ...
+    ignore_keys:
+    - cond_stage_model
+    linear_start: 0.0015
+    linear_end: 0.0195
+    timesteps: 1000
+    num_timesteps_cond: 1
+    first_stage_key: image
+    image_size: 16
+    channels: 3
+    cond_stage_trainable: false
+    concat_mode: false
+    scale_by_std: true
+    train_attention_only: true
+    attention_flag: self
+    monitor: val/loss_simple_ema
+    log_every_t: 200
+    dp_config:
+      enabled: true
+      epsilon: 10
+      delta: 1.0e-06
+      max_grad_norm: 0.0005
+      poisson_sampling: true
+      max_batch_size: 100
+    unet_config:
+      target: ldm.modules.diffusionmodules.openaimodel.UNetModel
+      params:
+        image_size: 16
+        in_channels: 3
+        out_channels: 3
+        model_channels: 192
+        attention_resolutions:
+        - 4
+        - 2
+        - 1
+        num_res_blocks: 2
+        channel_mult:
+        - 1
+        - 2
+        - 4
+        num_head_channels: 32
+        use_scale_shift_norm: true
+        resblock_updown: true
+    first_stage_config:
+      target: ldm.models.autoencoder.AutoencoderKL
+      params:
+        monitor: val/rec_loss
+        embed_dim: 3
+        lossconfig:
+          target: ldm.modules.losses.LPIPSWithDiscriminator
+          params:
+            disc_start: 50001
+            kl_weight: 1.0e-06
+            disc_weight: 0.5
+        ddconfig:
+          double_z: true
+          z_channels: 3
+          resolution: 32
+          in_channels: 3
+          out_ch: 3
+          ch: 128
+          ch_mult:
+          - 1
+          - 2
+          num_res_blocks: 2
+          attn_resolutions:
+          - 16
+          - 8
+          dropout: 0.0
+    cond_stage_config: __is_unconditional__
+    ablation_blocks: -1
+data:
+  target: main.DataModuleFromConfig
+  params:
+    batch_size: 8192
+    num_workers: 4
+    wrap: false
+    train:
+      target: ldm.data.celeba.CelebATrain
+      params:
+        config:
+          size: 32
+    validation:
+      target: ldm.data.celeba.CelebAValidation
+      params:
+        config:
+          size: 32
+lightning:
+  callbacks:
+    image_logger:
+      target: main.ImageLogger
+      params:
+        batch_frequency: 500
+        max_images: 8
+        increase_log_steps: false
+  trainer:
+    benchmark: true
+    max_epochs: 20
+    accelerator: gpu
+    gpus: 0,

celeba32/celeba32_eps5.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bfa5b3a832bd3412f2ec8cec73b43603a818b6be1a010ae49b86ec47094072af
+size 1570608639

celeba32/celeba32_eps5.yaml

@@ -0,0 +1,104 @@
+model:
+  base_learning_rate: 5.0e-07
+  target: ldm.models.diffusion.ddpm.LatentDiffusion
+  params:
+    ckpt_path: ...
+    ignore_keys:
+    - cond_stage_model
+    linear_start: 0.0015
+    linear_end: 0.0195
+    timesteps: 1000
+    num_timesteps_cond: 1
+    first_stage_key: image
+    image_size: 16
+    channels: 3
+    cond_stage_trainable: false
+    concat_mode: false
+    scale_by_std: true
+    train_attention_only: true
+    attention_flag: self
+    monitor: val/loss_simple_ema
+    log_every_t: 200
+    dp_config:
+      enabled: true
+      epsilon: 5
+      delta: 1.0e-06
+      max_grad_norm: 0.0005
+      poisson_sampling: true
+      max_batch_size: 100
+    unet_config:
+      target: ldm.modules.diffusionmodules.openaimodel.UNetModel
+      params:
+        image_size: 16
+        in_channels: 3
+        out_channels: 3
+        model_channels: 192
+        attention_resolutions:
+        - 4
+        - 2
+        - 1
+        num_res_blocks: 2
+        channel_mult:
+        - 1
+        - 2
+        - 4
+        num_head_channels: 32
+        use_scale_shift_norm: true
+        resblock_updown: true
+    first_stage_config:
+      target: ldm.models.autoencoder.AutoencoderKL
+      params:
+        monitor: val/rec_loss
+        embed_dim: 3
+        lossconfig:
+          target: ldm.modules.losses.LPIPSWithDiscriminator
+          params:
+            disc_start: 50001
+            kl_weight: 1.0e-06
+            disc_weight: 0.5
+        ddconfig:
+          double_z: true
+          z_channels: 3
+          resolution: 32
+          in_channels: 3
+          out_ch: 3
+          ch: 128
+          ch_mult:
+          - 1
+          - 2
+          num_res_blocks: 2
+          attn_resolutions:
+          - 16
+          - 8
+          dropout: 0.0
+    cond_stage_config: __is_unconditional__
+    ablation_blocks: -1
+data:
+  target: main.DataModuleFromConfig
+  params:
+    batch_size: 8192
+    num_workers: 4
+    wrap: false
+    train:
+      target: ldm.data.celeba.CelebATrain
+      params:
+        config:
+          size: 32
+    validation:
+      target: ldm.data.celeba.CelebAValidation
+      params:
+        config:
+          size: 32
+lightning:
+  callbacks:
+    image_logger:
+      target: main.ImageLogger
+      params:
+        batch_frequency: 500
+        max_images: 8
+        increase_log_steps: false
+  trainer:
+    benchmark: true
+    max_epochs: 20
+    accelerator: gpu
+    gpus: 0,

celeba64/celeba64_eps1.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fc76d82cd2afd88a0351c0d68cd92b68a0aa4eb712f5b138ecd98e9e9901ea92
+size 542413849

celeba64/celeba64_eps1.yaml

@@ -0,0 +1,103 @@
+model:
+  base_learning_rate: 1.0e-07
+  target: ldm.models.diffusion.ddpm.LatentDiffusion
+  params:
+    ckpt_path: ...
+    ignore_keys:
+    - cond_stage_model
+    - model_ema
+    linear_start: 0.0015
+    linear_end: 0.0195
+    timesteps: 1000
+    num_timesteps_cond: 1
+    first_stage_key: image
+    image_size: 32
+    channels: 3
+    cond_stage_trainable: false
+    train_attention_only: true
+    use_ema: false
+    monitor: val/loss
+    log_every_t: 200
+    dp_config:
+      enabled: true
+      epsilon: 1
+      delta: 1.0e-06
+      max_grad_norm: 0.0005
+      poisson_sampling: true
+      max_batch_size: 50
+    unet_config:
+      target: ldm.modules.diffusionmodules.openaimodel.UNetModel
+      params:
+        image_size: 32
+        in_channels: 3
+        out_channels: 3
+        model_channels: 128
+        attention_resolutions:
+        - 4
+        - 2
+        - 1
+        num_res_blocks: 2
+        channel_mult:
+        - 1
+        - 2
+        - 4
+        num_heads: 8
+        use_scale_shift_norm: true
+        resblock_updown: true
+    first_stage_config:
+      target: ldm.models.autoencoder.AutoencoderKL
+      params:
+        monitor: val/rec_loss
+        embed_dim: 3
+        lossconfig:
+          target: ldm.modules.losses.LPIPSWithDiscriminator
+          params:
+            disc_start: 50001
+            kl_weight: 1.0e-06
+            disc_weight: 0.5
+        ddconfig:
+          double_z: true
+          z_channels: 3
+          resolution: 64
+          in_channels: 3
+          out_ch: 3
+          ch: 192
+          ch_mult:
+          - 1
+          - 2
+          num_res_blocks: 2
+          attn_resolutions:
+          - 16
+          - 8
+          dropout: 0.0
+    cond_stage_config: __is_unconditional__
+    ablation_blocks: -1
+data:
+  target: main.DataModuleFromConfig
+  params:
+    batch_size: 8192
+    num_workers: 6
+    wrap: false
+    train:
+      target: ldm.data.celeba.CelebATrain
+      params:
+        config:
+          size: 64
+    validation:
+      target: ldm.data.celeba.CelebAValidation
+      params:
+        config:
+          size: 64
+lightning:
+  callbacks:
+    image_logger:
+      target: main.ImageLogger
+      params:
+        batch_frequency: 25000
+        max_images: 8
+        increase_log_steps: false
+  trainer:
+    benchmark: true
+    max_epochs: 70
+    accelerator: gpu
+    gpus: 0,

celeba64/celeba64_eps10.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b4b6875c0421b5308da4aa58087a95e8cdf0b29ef3fd8d5cd50d5be37f7063a5
+size 542413849

celeba64/celeba64_eps10.yaml

@@ -0,0 +1,103 @@
+model:
+  base_learning_rate: 1.0e-07
+  target: ldm.models.diffusion.ddpm.LatentDiffusion
+  params:
+    ckpt_path: ...
+    ignore_keys:
+    - cond_stage_model
+    - model_ema
+    linear_start: 0.0015
+    linear_end: 0.0195
+    timesteps: 1000
+    num_timesteps_cond: 1
+    first_stage_key: image
+    image_size: 32
+    channels: 3
+    cond_stage_trainable: false
+    train_attention_only: true
+    use_ema: false
+    monitor: val/loss
+    log_every_t: 200
+    dp_config:
+      enabled: true
+      epsilon: 10
+      delta: 1.0e-06
+      max_grad_norm: 0.0005
+      poisson_sampling: true
+      max_batch_size: 50
+    unet_config:
+      target: ldm.modules.diffusionmodules.openaimodel.UNetModel
+      params:
+        image_size: 32
+        in_channels: 3
+        out_channels: 3
+        model_channels: 128
+        attention_resolutions:
+        - 4
+        - 2
+        - 1
+        num_res_blocks: 2
+        channel_mult:
+        - 1
+        - 2
+        - 4
+        num_heads: 8
+        use_scale_shift_norm: true
+        resblock_updown: true
+    first_stage_config:
+      target: ldm.models.autoencoder.AutoencoderKL
+      params:
+        monitor: val/rec_loss
+        embed_dim: 3
+        lossconfig:
+          target: ldm.modules.losses.LPIPSWithDiscriminator
+          params:
+            disc_start: 50001
+            kl_weight: 1.0e-06
+            disc_weight: 0.5
+        ddconfig:
+          double_z: true
+          z_channels: 3
+          resolution: 64
+          in_channels: 3
+          out_ch: 3
+          ch: 192
+          ch_mult:
+          - 1
+          - 2
+          num_res_blocks: 2
+          attn_resolutions:
+          - 16
+          - 8
+          dropout: 0.0
+    cond_stage_config: __is_unconditional__
+    ablation_blocks: -1
+data:
+  target: main.DataModuleFromConfig
+  params:
+    batch_size: 8192
+    num_workers: 6
+    wrap: false
+    train:
+      target: ldm.data.celeba.CelebATrain
+      params:
+        config:
+          size: 64
+    validation:
+      target: ldm.data.celeba.CelebAValidation
+      params:
+        config:
+          size: 64
+lightning:
+  callbacks:
+    image_logger:
+      target: main.ImageLogger
+      params:
+        batch_frequency: 25000
+        max_images: 8
+        increase_log_steps: false
+  trainer:
+    benchmark: true
+    max_epochs: 70
+    accelerator: gpu
+    gpus: 0,

celeba64/celeba64_eps5.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6b13337d51f2602d8822b7cef7d6b043c1fa175c3ef38888b1ec8ce7d27e85d8
+size 542413849

celeba64/celeba64_eps5.yaml

@@ -0,0 +1,103 @@
+model:
+  base_learning_rate: 1.0e-07
+  target: ldm.models.diffusion.ddpm.LatentDiffusion
+  params:
+    ckpt_path: ...
+    ignore_keys:
+    - cond_stage_model
+    - model_ema
+    linear_start: 0.0015
+    linear_end: 0.0195
+    timesteps: 1000
+    num_timesteps_cond: 1
+    first_stage_key: image
+    image_size: 32
+    channels: 3
+    cond_stage_trainable: false
+    train_attention_only: true
+    use_ema: false
+    monitor: val/loss
+    log_every_t: 200
+    dp_config:
+      enabled: true
+      epsilon: 5
+      delta: 1.0e-06
+      max_grad_norm: 0.0005
+      poisson_sampling: true
+      max_batch_size: 50
+    unet_config:
+      target: ldm.modules.diffusionmodules.openaimodel.UNetModel
+      params:
+        image_size: 32
+        in_channels: 3
+        out_channels: 3
+        model_channels: 128
+        attention_resolutions:
+        - 4
+        - 2
+        - 1
+        num_res_blocks: 2
+        channel_mult:
+        - 1
+        - 2
+        - 4
+        num_heads: 8
+        use_scale_shift_norm: true
+        resblock_updown: true
+    first_stage_config:
+      target: ldm.models.autoencoder.AutoencoderKL
+      params:
+        monitor: val/rec_loss
+        embed_dim: 3
+        lossconfig:
+          target: ldm.modules.losses.LPIPSWithDiscriminator
+          params:
+            disc_start: 50001
+            kl_weight: 1.0e-06
+            disc_weight: 0.5
+        ddconfig:
+          double_z: true
+          z_channels: 3
+          resolution: 64
+          in_channels: 3
+          out_ch: 3
+          ch: 192
+          ch_mult:
+          - 1
+          - 2
+          num_res_blocks: 2
+          attn_resolutions:
+          - 16
+          - 8
+          dropout: 0.0
+    cond_stage_config: __is_unconditional__
+    ablation_blocks: -1
+data:
+  target: main.DataModuleFromConfig
+  params:
+    batch_size: 8192
+    num_workers: 6
+    wrap: false
+    train:
+      target: ldm.data.celeba.CelebATrain
+      params:
+        config:
+          size: 64
+    validation:
+      target: ldm.data.celeba.CelebAValidation
+      params:
+        config:
+          size: 64
+lightning:
+  callbacks:
+    image_logger:
+      target: main.ImageLogger
+      params:
+        batch_frequency: 25000
+        max_images: 8
+        increase_log_steps: false
+  trainer:
+    benchmark: true
+    max_epochs: 70
+    accelerator: gpu
+    gpus: 0,