JosephusCheung/RuminationDiffusion

1model:2  base_learning_rate: 1.0e-43  target: ldm.models.diffusion.ddpm.LatentDiffusion4  params:5    parameterization: "v"6    linear_start: 0.000857    linear_end: 0.01208    num_timesteps_cond: 19    log_every_t: 20010    timesteps: 100011    first_stage_key: "jpg"12    cond_stage_key: "txt"13    image_size: 6414    channels: 415    cond_stage_trainable: false16    conditioning_key: crossattn17    monitor: val/loss_simple_ema18    scale_factor: 0.1821519    use_ema: False # we set this to false because this is an inference only config20 21    unet_config:22      target: ldm.modules.diffusionmodules.openaimodel.UNetModel23      params:24        use_checkpoint: True25        use_fp16: True26        image_size: 32 # unused27        in_channels: 428        out_channels: 429        model_channels: 32030        attention_resolutions: [ 4, 2, 1 ]31        num_res_blocks: 232        channel_mult: [ 1, 2, 4, 4 ]33        num_head_channels: 64 # need to fix for flash-attn34        use_spatial_transformer: True35        use_linear_in_transformer: True36        transformer_depth: 137        context_dim: 102438        legacy: False39 40    first_stage_config:41      target: ldm.models.autoencoder.AutoencoderKL42      params:43        embed_dim: 444        monitor: val/rec_loss45        ddconfig:46          #attn_type: "vanilla-xformers"47          double_z: true48          z_channels: 449          resolution: 25650          in_channels: 351          out_ch: 352          ch: 12853          ch_mult:54          - 155          - 256          - 457          - 458          num_res_blocks: 259          attn_resolutions: []60          dropout: 0.061        lossconfig:62          target: torch.nn.Identity63 64    cond_stage_config:65      target: ldm.modules.encoders.modules.FrozenOpenCLIPEmbedder66      params:67        freeze: True68        layer: "penultimate"69