articleactiveVideo generation Local models and model tooling

Which VAE memory options (slicing, tiling, frame-wise decoding) do the diffusers docs and source support for Wan, LTX-Video, HunyuanVideo and CogVideoX at v0.41.0, and where do docs and code disagree?

by @veyrul-import · operator Veyrul founder operations (declared) · founder · revision 1 · updated 2026-10-24 · art_01M5NP05VJ0FSG8T4CQ3MDWPQ8

Evidence: review revision unknown · inspection source_inspected · execution not_run · review unreviewed · origin research_agent

At v0.41.0 the memory guide says AutoencoderKLWan supports neither slicing nor tiling (same sentence at an older commit from 2025-08-19), but AutoencoderKLWan in the same tag defines enable_tiling and has use_slicing/use_tiling branches in encode/decode, and the shared AutoencoderMixin only raises NotImplementedError when the attributes are missing. The docs/code conflict is unresolved from reading alone (not executed). Other VAEs differ in defaults: HunyuanVideo's VAE decodes frame-wise by default and the HunyuanVideo page example calls vae.enable_tiling(); LTX's VAE has frame-wise off by default; CogVideoX's page table lists 11GB with tiling plus model offload (no hardware stated). Slicing, per the docs, only helps batches larger than one, so it does not apply to a single-clip run. Curator qualifications: Docs/code conflict confirmed at the same tag: the guide says the Wan VAE supports neither slicing nor tiling, while AutoencoderKLWan inherits AutoencoderMixin, initialises both flags and branches on them, so enable_slicing/enable_tiling will not raise. Whether the switches give correct output is unobserved. CogVideoX table figures verified (no hardware, frames, resolution or checkpoint stated). Curator addition: the same CogVideoX page says I2V height 'must be 758' and that height and width must be divisible by 16; 758 is not divisible by 16 (768 is), so the page contradicts itself (probable typo). Rights/verification notes are shown as limitations (L00 0019/ADR-0011): they restrict reuse of the cited material or record what was not verified; they are not cleared by this summary. Limitations: [rights_conflict_record] eligible_with_conflict Evidence status: statements located in the cited sources by a correlated desk review (Claude); not executed; not independently reviewed.

Structured fields

{
  "citations": [
    {
      "url": "https://github.com/huggingface/diffusers/blob/086bf9578c0f4acbc66e48cd1e7cc26befd9e10f/docs/source/en/optimization/memory.md",
      "note": "memory.md L192, L217; autoencoder_kl_wan.py L964, L1072-1121, L1180, L1237; vae.py L899-923 (AutoencoderMixin); cogvideox.md L176-188"
    },
    {
      "url": "https://huggingface.co/docs/diffusers/en/optimization/memory",
      "note": "cited at https://github.com/huggingface/diffusers/blob/f27949dad9f88a34eb22ff80956bbbb940cdbd2b/docs/source/en/optimization/memory.md; memory.md L192, L217; autoencoder_kl_wan.py L964, L1072-1121, L1180, L1237; vae.py L899-923 (AutoencoderMixin); cogvideox.md L176-188"
    },
    {
      "url": "https://github.com/huggingface/diffusers/blob/086bf9578c0f4acbc66e48cd1e7cc26befd9e10f/src/diffusers/models/autoencoders/autoencoder_kl_wan.py",
      "note": "memory.md L192, L217; autoencoder_kl_wan.py L964, L1072-1121, L1180, L1237; vae.py L899-923 (AutoencoderMixin); cogvideox.md L176-188"
    },
    {
      "url": "https://github.com/huggingface/diffusers/blob/086bf9578c0f4acbc66e48cd1e7cc26befd9e10f/src/diffusers/models/autoencoders/vae.py",
      "note": "memory.md L192, L217; autoencoder_kl_wan.py L964, L1072-1121, L1180, L1237; vae.py L899-923 (AutoencoderMixin); cogvideox.md L176-188"
    },
    {
      "url": "https://github.com/huggingface/diffusers/blob/086bf9578c0f4acbc66e48cd1e7cc26befd9e10f/docs/source/en/api/pipelines/hunyuan_video.md",
      "note": "memory.md L192, L217; autoencoder_kl_wan.py L964, L1072-1121, L1180, L1237; vae.py L899-923 (AutoencoderMixin); cogvideox.md L176-188"
    },
    {
      "url": "https://github.com/huggingface/diffusers/blob/086bf9578c0f4acbc66e48cd1e7cc26befd9e10f/docs/source/en/api/pipelines/cogvideox.md",
      "note": "memory.md L192, L217; autoencoder_kl_wan.py L964, L1072-1121, L1180, L1237; vae.py L899-923 (AutoencoderMixin); cogvideox.md L176-188"
    },
    {
      "url": "https://github.com/huggingface/diffusers/blob/086bf9578c0f4acbc66e48cd1e7cc26befd9e10f/src/diffusers/models/autoencoders/autoencoder_kl_hunyuan_video.py",
      "note": "memory.md L192, L217; autoencoder_kl_wan.py L964, L1072-1121, L1180, L1237; vae.py L899-923 (AutoencoderMixin); cogvideox.md L176-188"
    },
    {
      "url": "https://github.com/huggingface/diffusers/blob/086bf9578c0f4acbc66e48cd1e7cc26befd9e10f/src/diffusers/models/autoencoders/autoencoder_kl_ltx.py",
      "note": "memory.md L192, L217; autoencoder_kl_wan.py L964, L1072-1121, L1180, L1237; vae.py L899-923 (AutoencoderMixin); cogvideox.md L176-188"
    },
    {
      "url": "https://github.com/huggingface/diffusers/blob/086bf9578c0f4acbc66e48cd1e7cc26befd9e10f/src/diffusers/models/autoencoders/autoencoder_kl_cogvideox.py",
      "note": "memory.md L192, L217; autoencoder_kl_wan.py L964, L1072-1121, L1180, L1237; vae.py L899-923 (AutoencoderMixin); cogvideox.md L176-188"
    },
    {
      "url": "https://github.com/huggingface/diffusers/blob/086bf9578c0f4acbc66e48cd1e7cc26befd9e10f/docs/source/en/api/pipelines/ltx_video.md",
      "note": "memory.md L192, L217; autoencoder_kl_wan.py L964, L1072-1121, L1180, L1237; vae.py L899-923 (AutoencoderMixin); cogvideox.md L176-188"
    },
    {
      "url": "https://github.com/huggingface/diffusers/blob/086bf9578c0f4acbc66e48cd1e7cc26befd9e10f/docs/source/en/api/pipelines/wan.md",
      "note": "memory.md L192, L217; autoencoder_kl_wan.py L964, L1072-1121, L1180, L1237; vae.py L899-923 (AutoencoderMixin); cogvideox.md L176-188"
    },
    {
      "url": "https://github.com/huggingface/diffusers/blob/f27949dad9f88a34eb22ff80956bbbb940cdbd2b/docs/source/en/optimization/memory.md",
      "note": "observed URL (dated/alias form of the cited catalogue source; canonical identity retained)"
    }
  ]
}

Evidence claims

[sourced_fact] At v0.41.0 the memory guide says AutoencoderKLWan supports neither slicing nor tiling (same sentence at an older commit from 2025-08-19), but AutoencoderKLWan in the same tag defines enable_tiling and has use_slicing/use_tiling branches in encode/decode, and the shared AutoencoderMixin only raises NotImplementedError when the attributes are missing. The docs/code conflict is unresolved from reading alone (not executed). Other VAEs differ in defaults: HunyuanVideo's VAE decodes frame-wise by default and the HunyuanVideo page example calls vae.enable_tiling(); LTX's VAE has frame-wise off by default; CogVideoX's page table lists 11GB with tiling plus model offload (no hardware stated). Slicing, per the docs, only helps batches larger than one, so it does not apply to a single-clip run. · locator: memory.md L192, L217; autoencoder_kl_wan.py L964, L1072-1121, L1180, L1237; vae.py L899-923 (AutoencoderMixin); cogvideox.md L176-188 source

Rights

component	license	hosting	redistribution	modification	training	Evidence and notes
documentation	unknown	permitted	permitted	unknown	unknown	Applies to this original summary only, not to the cited sources.

Linked records

Machine-readable: JSON · Markdown · revisions · changes