fastvideo.pipelines.stages.image_encoding

Image encoding stages for I2V diffusion pipelines.

This module contains implementations of image encoding stages for diffusion pipelines.

Module Contents

Classes

ImageEncodingStage	Stage for encoding image prompts into embeddings for diffusion models.
ImageVAEEncodingStage	Stage for encoding pixel representations into latent space.

Data

logger

API

class fastvideo.pipelines.stages.image_encoding.ImageEncodingStage(image_encoder, image_processor)

Bases: fastvideo.pipelines.stages.base.PipelineStage

Stage for encoding image prompts into embeddings for diffusion models.

This stage handles the encoding of image prompts into the embedding space expected by the diffusion model.

Initialization

Initialize the prompt encoding stage.

Parameters:

• enable_logging – Whether to enable logging for this stage.

• is_secondary – Whether this is a secondary image encoder.

forward(batch: fastvideo.pipelines.pipeline_batch_info.ForwardBatch, fastvideo_args: fastvideo.fastvideo_args.FastVideoArgs) → fastvideo.pipelines.pipeline_batch_info.ForwardBatch

Encode the prompt into image encoder hidden states.

Parameters:

• batch – The current batch information.

• fastvideo_args – The inference arguments.

Returns:

The batch with encoded prompt embeddings.

verify_input(batch: fastvideo.pipelines.pipeline_batch_info.ForwardBatch, fastvideo_args: fastvideo.fastvideo_args.FastVideoArgs) → fastvideo.pipelines.stages.validators.VerificationResult

Verify image encoding stage inputs.

verify_output(batch: fastvideo.pipelines.pipeline_batch_info.ForwardBatch, fastvideo_args: fastvideo.fastvideo_args.FastVideoArgs) → fastvideo.pipelines.stages.validators.VerificationResult

Verify image encoding stage outputs.

class fastvideo.pipelines.stages.image_encoding.ImageVAEEncodingStage(vae: fastvideo.models.vaes.common.ParallelTiledVAE)

Bases: fastvideo.pipelines.stages.base.PipelineStage

Stage for encoding pixel representations into latent space.

This stage handles the encoding of pixel representations into the final input format (e.g., latents).

Initialization

forward(batch: fastvideo.pipelines.pipeline_batch_info.ForwardBatch, fastvideo_args: fastvideo.fastvideo_args.FastVideoArgs) → fastvideo.pipelines.pipeline_batch_info.ForwardBatch

Encode pixel representations into latent space.

Parameters:

• batch – The current batch information.

• fastvideo_args – The inference arguments.

Returns:

The batch with encoded outputs.

preprocess(image: torch.Tensor | PIL.Image.Image, vae_scale_factor: int, height: int | None = None, width: int | None = None, resize_mode: str = 'default') → torch.Tensor

retrieve_latents(encoder_output: torch.Tensor, generator: torch.Generator | None = None, sample_mode: str = 'sample')

verify_input(batch: fastvideo.pipelines.pipeline_batch_info.ForwardBatch, fastvideo_args: fastvideo.fastvideo_args.FastVideoArgs) → fastvideo.pipelines.stages.validators.VerificationResult

Verify encoding stage inputs.

verify_output(batch: fastvideo.pipelines.pipeline_batch_info.ForwardBatch, fastvideo_args: fastvideo.fastvideo_args.FastVideoArgs) → fastvideo.pipelines.stages.validators.VerificationResult

Verify encoding stage outputs.

fastvideo.pipelines.stages.image_encoding.logger

‘init_logger(…)’