Repository navigation
fix(video): Wan 2.2 5B offloads on 24 GiB cards - #307
Merged
Merged
Conversation
Loading the whole 5B pipeline (fp16 transformer + bf16 UMT5 + fp32 VAE, ~24.2 GB) onto an RTX 3090 left it at 0/30 steps for 10+ minutes on Salad: under WSL the driver spills to system memory instead of raising OOM. With full_vram_gib a 24 GiB card encodes the prompt on the CPU and uses model offload + VAE tiling, like Wan 2.1 1.3B on 8 GiB. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
ThreatCrush Security Scan13 finding(s) HIGH/CRITICAL: 1 | MEDIUM: 8 | LOW: 4
Snippets are redacted; ThreatCrush never prints matched credential material. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
The 5B pipeline is ~24.2 GB fully loaded (fp16 transformer + bf16 UMT5 + fp32 VAE). On a Salad RTX 3090 it sat at 0/30 steps for 10+ min: WSL spills VRAM to system memory instead of OOMing.
full_vram_gib = 28makes 24 GiB cards encode the prompt on CPU and run model offload + VAE tiling. Contract tests updated.🤖 Generated with Claude Code