Skip to content
代码片段 群组 项目
选择Git版本
  • 6ff2e4d55002cca7ed836dfb39fdf0c82a7f615a
  • master 默认 受保护
  • flipflop-stream
  • asset-management
  • cache-index-json-locales
  • expand_sampler_cfg_function
  • worksplit-multigpu
  • dd
  • diffdiff-strength
  • error-catch-fix
  • attention-select
  • fix-context-window-slicing
  • combo-output
  • clarify-io-typehint
  • fix-easycache-logs
  • js/progress-crossover-fix-clean
  • js/progress-crossover-fix
  • sortblock
  • gemini-image
  • easycache
  • node-memory-reserve
  • v0.3.64
  • v0.3.63
  • v0.3.62
  • v0.3.61
  • v0.3.60
  • v0.3.59
  • v0.3.58
  • v0.3.57
  • v0.3.56
  • v0.3.55
  • v0.3.54
  • v0.3.53
  • v0.3.52
  • v0.3.51
  • v0.3.50
  • v0.3.49
  • v0.3.48
  • v0.3.47
  • v0.3.46
  • v0.3.45
41 个结果

latent_preview.py

Blame
  • 代码所有者
    将用户和群组指定为特定文件更改的核准人。 了解更多。
    latent_preview.py 4.25 KiB
    import torch
    from PIL import Image
    from comfy.cli_args import args, LatentPreviewMethod
    from comfy.taesd.taesd import TAESD
    import comfy.model_management
    import folder_paths
    import comfy.utils
    import logging
    
    MAX_PREVIEW_RESOLUTION = args.preview_size
    
    def preview_to_image(latent_image):
            latents_ubyte = (((latent_image + 1.0) / 2.0).clamp(0, 1)  # change scale from -1..1 to 0..1
                                .mul(0xFF)  # to 0..255
                                )
            if comfy.model_management.directml_enabled:
                    latents_ubyte = latents_ubyte.to(dtype=torch.uint8)
            latents_ubyte = latents_ubyte.to(device="cpu", dtype=torch.uint8, non_blocking=comfy.model_management.device_supports_non_blocking(latent_image.device))
    
            return Image.fromarray(latents_ubyte.numpy())
    
    class LatentPreviewer:
        def decode_latent_to_preview(self, x0):
            pass
    
        def decode_latent_to_preview_image(self, preview_format, x0):
            preview_image = self.decode_latent_to_preview(x0)
            return ("JPEG", preview_image, MAX_PREVIEW_RESOLUTION)
    
    class TAESDPreviewerImpl(LatentPreviewer):
        def __init__(self, taesd):
            self.taesd = taesd
    
        def decode_latent_to_preview(self, x0):
            x_sample = self.taesd.decode(x0[:1])[0].movedim(0, 2)
            return preview_to_image(x_sample)
    
    
    class Latent2RGBPreviewer(LatentPreviewer):
        def __init__(self, latent_rgb_factors, latent_rgb_factors_bias=None):
            self.latent_rgb_factors = torch.tensor(latent_rgb_factors, device="cpu").transpose(0, 1)
            self.latent_rgb_factors_bias = None
            if latent_rgb_factors_bias is not None:
                self.latent_rgb_factors_bias = torch.tensor(latent_rgb_factors_bias, device="cpu")
    
        def decode_latent_to_preview(self, x0):
            self.latent_rgb_factors = self.latent_rgb_factors.to(dtype=x0.dtype, device=x0.device)
            if self.latent_rgb_factors_bias is not None:
                self.latent_rgb_factors_bias = self.latent_rgb_factors_bias.to(dtype=x0.dtype, device=x0.device)
    
            if x0.ndim == 5:
                x0 = x0[0, :, 0]
            else:
                x0 = x0[0]
    
            latent_image = torch.nn.functional.linear(x0.movedim(0, -1), self.latent_rgb_factors, bias=self.latent_rgb_factors_bias)
            # latent_image = x0[0].permute(1, 2, 0) @ self.latent_rgb_factors
    
            return preview_to_image(latent_image)
    
    
    def get_previewer(device, latent_format):
        previewer = None
        method = args.preview_method
        if method != LatentPreviewMethod.NoPreviews:
            # TODO previewer methods
            taesd_decoder_path = None
            if latent_format.taesd_decoder_name is not None:
                taesd_decoder_path = next(
                    (fn for fn in folder_paths.get_filename_list("vae_approx")
                        if fn.startswith(latent_format.taesd_decoder_name)),
                    ""
                )
                taesd_decoder_path = folder_paths.get_full_path("vae_approx", taesd_decoder_path)
    
            if method == LatentPreviewMethod.Auto:
                method = LatentPreviewMethod.Latent2RGB
    
            if method == LatentPreviewMethod.TAESD:
                if taesd_decoder_path:
                    taesd = TAESD(None, taesd_decoder_path, latent_channels=latent_format.latent_channels).to(device)
                    previewer = TAESDPreviewerImpl(taesd)
                else:
                    logging.warning("Warning: TAESD previews enabled, but could not find models/vae_approx/{}".format(latent_format.taesd_decoder_name))
    
            if previewer is None:
                if latent_format.latent_rgb_factors is not None:
                    previewer = Latent2RGBPreviewer(latent_format.latent_rgb_factors, latent_format.latent_rgb_factors_bias)
        return previewer
    
    def prepare_callback(model, steps, x0_output_dict=None):
        preview_format = "JPEG"
        if preview_format not in ["JPEG", "PNG"]:
            preview_format = "JPEG"
    
        previewer = get_previewer(model.load_device, model.model.latent_format)
    
        pbar = comfy.utils.ProgressBar(steps)
        def callback(step, x0, x, total_steps):
            if x0_output_dict is not None:
                x0_output_dict["x0"] = x0
    
            preview_bytes = None
            if previewer:
                preview_bytes = previewer.decode_latent_to_preview_image(preview_format, x0)
            pbar.update_absolute(step + 1, total_steps, preview_bytes)
        return callback