PeggyWang/future-diffusion

1model:2  base_learning_rate: 1.0e-43  target: ldm.models.diffusion.ddpm.LatentDiffusion4  params:5    linear_start: 0.000856    linear_end: 0.01207    num_timesteps_cond: 18    log_every_t: 2009    timesteps: 100010    first_stage_key: "jpg"11    cond_stage_key: "txt"12    image_size: 6413    channels: 414    cond_stage_trainable: false15    conditioning_key: crossattn16    monitor: val/loss_simple_ema17    scale_factor: 0.1821518    use_ema: False # we set this to false because this is an inference only config19 20    unet_config:21      target: ldm.modules.diffusionmodules.openaimodel.UNetModel22      params:23        use_checkpoint: True24        use_fp16: True25        image_size: 32 # unused26        in_channels: 427        out_channels: 428        model_channels: 32029        attention_resolutions: [ 4, 2, 1 ]30        num_res_blocks: 231        channel_mult: [ 1, 2, 4, 4 ]32        num_head_channels: 64 # need to fix for flash-attn33        use_spatial_transformer: True34        use_linear_in_transformer: True35        transformer_depth: 136        context_dim: 102437        legacy: False38 39    first_stage_config:40      target: ldm.models.autoencoder.AutoencoderKL41      params:42        embed_dim: 443        monitor: val/rec_loss44        ddconfig:45          #attn_type: "vanilla-xformers"46          double_z: true47          z_channels: 448          resolution: 25649          in_channels: 350          out_ch: 351          ch: 12852          ch_mult:53          - 154          - 255          - 456          - 457          num_res_blocks: 258          attn_resolutions: []59          dropout: 0.060        lossconfig:61          target: torch.nn.Identity62 63    cond_stage_config:64      target: ldm.modules.encoders.modules.FrozenOpenCLIPEmbedder65      params:66        freeze: True67        layer: "penultimate"68