quantumcontrol/stable-video-diffusion

1model:2  target: sgm.models.diffusion.DiffusionEngine3  params:4    scale_factor: 0.182155    disable_first_stage_autocast: True6    ckpt_path: checkpoints/svd_xt.safetensors7 8    denoiser_config:9      target: sgm.modules.diffusionmodules.denoiser.Denoiser10      params:11        scaling_config:12          target: sgm.modules.diffusionmodules.denoiser_scaling.VScalingWithEDMcNoise13 14    network_config:15      target: sgm.modules.diffusionmodules.video_model.VideoUNet16      params:17        adm_in_channels: 76818        num_classes: sequential19        use_checkpoint: True20        in_channels: 821        out_channels: 422        model_channels: 32023        attention_resolutions: [4, 2, 1]24        num_res_blocks: 225        channel_mult: [1, 2, 4, 4]26        num_head_channels: 6427        use_linear_in_transformer: True28        transformer_depth: 129        context_dim: 102430        spatial_transformer_attn_type: softmax-xformers31        extra_ff_mix_layer: True32        use_spatial_context: True33        merge_strategy: learned_with_images34        video_kernel_size: [3, 1, 1]35 36    conditioner_config:37      target: sgm.modules.GeneralConditioner38      params:39        emb_models:40        - is_trainable: False41          input_key: cond_frames_without_noise42          target: sgm.modules.encoders.modules.FrozenOpenCLIPImagePredictionEmbedder43          params:44            n_cond_frames: 145            n_copies: 146            open_clip_embedding_config:47              target: sgm.modules.encoders.modules.FrozenOpenCLIPImageEmbedder48              params:49                freeze: True50 51        - input_key: fps_id52          is_trainable: False53          target: sgm.modules.encoders.modules.ConcatTimestepEmbedderND54          params:55            outdim: 25656 57        - input_key: motion_bucket_id58          is_trainable: False59          target: sgm.modules.encoders.modules.ConcatTimestepEmbedderND60          params:61            outdim: 25662 63        - input_key: cond_frames64          is_trainable: False65          target: sgm.modules.encoders.modules.VideoPredictionEmbedderWithEncoder66          params:67            disable_encoder_autocast: True68            n_cond_frames: 169            n_copies: 170            is_ae: True71            encoder_config:72              target: sgm.models.autoencoder.AutoencoderKLModeOnly73              params:74                embed_dim: 475                monitor: val/rec_loss76                ddconfig:77                  attn_type: vanilla-xformers78                  double_z: True79                  z_channels: 480                  resolution: 25681                  in_channels: 382                  out_ch: 383                  ch: 12884                  ch_mult: [1, 2, 4, 4]85                  num_res_blocks: 286                  attn_resolutions: []87                  dropout: 0.088                lossconfig:89                  target: torch.nn.Identity90 91        - input_key: cond_aug92          is_trainable: False93          target: sgm.modules.encoders.modules.ConcatTimestepEmbedderND94          params:95            outdim: 25696 97    first_stage_config:98      target: sgm.models.autoencoder.AutoencodingEngine99      params:100        loss_config:101          target: torch.nn.Identity102        regularizer_config:103          target: sgm.modules.autoencoding.regularizers.DiagonalGaussianRegularizer104        encoder_config: 105          target: sgm.modules.diffusionmodules.model.Encoder106          params:107            attn_type: vanilla108            double_z: True109            z_channels: 4110            resolution: 256111            in_channels: 3112            out_ch: 3113            ch: 128114            ch_mult: [1, 2, 4, 4]115            num_res_blocks: 2116            attn_resolutions: []117            dropout: 0.0118        decoder_config:119          target: sgm.modules.autoencoding.temporal_ae.VideoDecoder120          params:121            attn_type: vanilla122            double_z: True123            z_channels: 4124            resolution: 256125            in_channels: 3126            out_ch: 3127            ch: 128128            ch_mult: [1, 2, 4, 4]129            num_res_blocks: 2130            attn_resolutions: []131            dropout: 0.0132            video_kernel_size: [3, 1, 1]133 134    sampler_config:135      target: sgm.modules.diffusionmodules.sampling.EulerEDMSampler136      params:137        discretization_config:138          target: sgm.modules.diffusionmodules.discretizer.EDMDiscretization139          params:140            sigma_max: 700.0141 142        guider_config:143          target: sgm.modules.diffusionmodules.guiders.LinearPredictionGuider144          params:145            max_scale: 3.0146            min_scale: 1.5