Upload v1-finetune_everydream.yaml

v1-finetune_everydream.yaml

@@ -0,0 +1,108 @@
+model:
+  base_learning_rate: 1.0e-6
+  target: ldm.models.diffusion.ddpm.LatentDiffusion
+  params:
+    linear_start: 0.00085
+    linear_end: 0.0120
+    num_timesteps_cond: 1
+    log_every_t: 200
+    timesteps: 1000
+    first_stage_key: image
+    cond_stage_key: caption
+    image_size: 64
+    channels: 4
+    cond_stage_trainable: true
+    conditioning_key: crossattn
+    monitor: val/loss_simple_ema
+    scale_factor: 0.18215
+    use_ema: False
+    unfreeze_model: True
+    model_lr: 1.0e-6
+
+    unet_config:
+      target: ldm.modules.diffusionmodules.openaimodel.UNetModel
+      params:
+        image_size: 32 # unused
+        in_channels: 4
+        out_channels: 4
+        model_channels: 320
+        attention_resolutions: [ 4, 2, 1 ]
+        num_res_blocks: 2
+        channel_mult: [ 1, 2, 4, 4 ]
+        num_heads: 8
+        use_spatial_transformer: True
+        transformer_depth: 1
+        context_dim: 768
+        use_checkpoint: True
+        legacy: False
+
+    first_stage_config:
+      target: ldm.models.autoencoder.AutoencoderKL
+      params:
+        embed_dim: 4
+        monitor: val/rec_loss
+        ddconfig:
+          double_z: true
+          z_channels: 4
+          resolution: 512
+          in_channels: 3
+          out_ch: 3
+          ch: 128
+          ch_mult:
+          - 1
+          - 2
+          - 4
+          - 4
+          num_res_blocks: 2
+          attn_resolutions: []
+          dropout: 0.0
+        lossconfig:
+          target: torch.nn.Identity
+
+    cond_stage_config:
+      target: ldm.modules.encoders.modules.FrozenCLIPEmbedder
+
+data:
+  target: main.DataModuleFromConfig
+  params:
+    batch_size: 4  # prefer highest possible without getting CUDA Out of Memory error, A100 40GB =~20 80GB= ~48
+    num_workers: 8
+    wrap: falsegit
+    train:
+      target: ldm.data.every_dream.EveryDreamBatch
+      params:
+        repeats: 5   # rough suggestions: 5 with 5000+ images, 15 for 1000 images, use micro yaml for <100
+        debug_level: 1   # 1 to print if images are dropped due to multiple-aspect ratio image batching
+        conditional_dropout: 0.04   # experimental, likelihood to drop the caption, may help with poorly captioned images
+        resolution: 512   # use 512 for 24GB, can use 576, 640, 704, 768, on higher VRAM cards only..
+    validation:
+      target: ldm.data.ed_validate.EDValidateBatch
+      params:
+        repeats: 0.5
+    test:
+      target: ldm.data.ed_validate.EDValidateBatch
+      params:
+        repeats: 0.2
+
+lightning:
+  modelcheckpoint:
+    params:
+      every_n_epochs: 1  # produce a ckpt every epoch, leave 1!
+      #every_n_train_steps: 1400 # can only use epoch or train step checkpoints
+      save_top_k: 6   # save the best N ckpts according to loss, can reduce to save disk space but suggest at LEAST 2, more if you have max_epochs below higher!
+      save_last: True
+      filename: "{epoch:02d}-{step:05d}"
+  callbacks:
+    image_logger:
+      target: main.ImageLogger
+      params:
+        batch_frequency: 200
+        max_images: 16
+        increase_log_steps: False
+
+  trainer:
+    benchmark: True
+    max_epochs: 6   # better to run several epochs and test your checkpoints!  Try 4-5, you get a checkpoint every epoch to test! 
+    max_steps: 99000   # better to end on epochs not steps, especially with >500 images to ensure even distribution, but you can set this if you really want...
+    check_val_every_n_epoch: 1
+    gpus: 0,