NodeValid

Published video-model weights, licences and hardware notes

Pyramid Flow discusses offloading without figures

Offloading is discussed on this card and no gigabyte figure appears anywhere on it. The reader is told which lever exists and not what pulling it costs or saves, so the hardware column stays empty. As of 2026-09-22.

Pyramid Flow in the hardware column, as publishedNaming a memory technique without a figure tells a reader that memory was a consideration and leaves them no way to check their own card against it.The memory statement as publishedNo memory figure recordedOffloading is discussed without gigabytes768p, with a 384pcheckpointStated per checkpoint10 seconds at 768p, 5 at384pSeconds, at 24 FPS, for both checkpointsA technique named, an effect unstated
Fig. 1 The lever is named and never measured, so the column records an absence.
What the repositories actually stateFilled where the repository states something, hollow where it does not. A blank is not an inference.What the repositories actually stateWhat the pages stateWhat the pagesstateHardware, in the vendor's own wordsHardware, in the vendor's own words — What the pages state: No memory figure recordedThe detail published beside itThe detail published beside it — What the pages state: Offloading is discussed without gigabytesResolution, from the same pagesResolution, from the same pages — What the pages state: 768p, with a 384p checkpointLength, from the same pagesLength, from the same pages — What the pages state: 10 seconds at 768p, 5 at 384p
Fig. 2 Filled where the repository states something, hollow where it does not. A blank is not an inference.
Pyramid Flow in the hardware column, with the columns that qualify it. Read from huggingface.co/rain1011 on 2026-09-22.
ColumnWhat the pages state
Hardware, in the vendor's own wordsNo memory figure recorded
The detail published beside itOffloading is discussed without gigabytes
Resolution, from the same pages768p, with a 384p checkpoint
Length, from the same pages10 seconds at 768p, 5 at 384p

Inclusion rule. The value this page is about, the detail the vendor attached to it, and the two columns that change how it should be read. A column the vendor left empty keeps its row and says so. Order. This column first, then its detail, then the two columns that qualify it.

1What mentioning offloading implies

Nobody documents offloading for a model that fits comfortably. The mention is itself evidence that memory is the binding constraint, which is a weak signal and a real one.

Another release in the register publishes both sides of the same trade, 9.3G with offload against 27.5G without. That is what this card would have looked like with numbers.

2The card is detailed about output

Ten seconds at 768p and five seconds at 384p, both at 24 FPS, per checkpoint. The resolution and length columns are among the better filled in the register.

Documentation effort went to what comes out rather than to what it runs on. That is a coherent choice for a research release and an awkward one for anyone buying hardware.

3Why no figure is estimated here

The card also publishes no parameter count, so even the unsafe arithmetic is unavailable. There is nothing on the page to estimate from.

The register records the absence and names the technique the card does discuss. A number from a forum thread would be the one value on this page nobody published.

4Sources

Every figure above is the vendor's own, taken from huggingface.co/rain1011 on 2026-09-22 with nothing measured here. The rest of the release is on its family entry, the whole column is on hardware, and all memory figures together are on requirements. Nearby: 14.7GB or 51.2GB, Eight H100s or one 4090.