Skip to content

Fix: Avoid crash and warn when using Qwen 2D VAE for Wan video #1721

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
leejet merged 2 commits into from
Jun 29, 2026
+16 −1
Merged

Fix: Avoid crash and warn when using Qwen 2D VAE for Wan video #1721

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
6e7bef1
Fix: Avoid crash and warn when using 2D VAE for video
stduhpf Jun 29, 2026
b86dd74
format
stduhpf Jun 29, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
17 changes: 16 additions & 1 deletion src/model/vae/wan_vae.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1119,6 +1119,10 @@ namespace WAN {
GGML_ASSERT(b == 1);
GGML_ASSERT(decode_only == false);

if (x->ne[2] > 1 && is_2D) {
LOG_WARN("Using 2D VAE to encode video, expect poor results");
}

clear_cache();

if (wan2_2) {
Expand All @@ -1138,7 +1142,8 @@ namespace WAN {
auto in = ggml_ext_slice(ctx->ggml_ctx, x, 2, 0, 1); // [b*c, 1, h, w]
out = encoder->forward(ctx, in, b, _enc_feat_map, _enc_conv_idx, i);
} else {
auto in = ggml_ext_slice(ctx->ggml_ctx, x, 2, 1 + 4 * (i - 1), 1 + 4 * i); // [b*c, 4, h, w]
// if is_2D, drop 3 out of 4 frames
auto in = ggml_ext_slice(ctx->ggml_ctx, x, 2, 1 + 4 * (i - 1), (is_2D ? 1 - 3 : 1) + 4 * i); // [b*c, 4, h, w]
auto out_ = encoder->forward(ctx, in, b, _enc_feat_map, _enc_conv_idx, i);
out = ggml_concat(ctx->ggml_ctx, out, out_, 2);
}
Expand All @@ -1161,6 +1166,10 @@ namespace WAN {
// z: [b*c, t, h, w]
GGML_ASSERT(b == 1);

if (z->ne[2] > 1 && is_2D) {
LOG_WARN("Using 2D VAE to decode video, expect poor results");
}

clear_cache();

auto decoder = std::dynamic_pointer_cast<Decoder3d>(blocks["decoder"]);
Expand All @@ -1185,6 +1194,12 @@ namespace WAN {
auto in = ggml_ext_slice(ctx->ggml_ctx, x, 2, i, i + 1); // [b*c, 1, h, w]
auto out_ = decoder->forward(ctx, in, b, _feat_map, _conv_idx, i);
out = ggml_concat(ctx->ggml_ctx, out, out_, 2);
if (is_2D) {
// repeat frames to avoid mismatch
for (int j = 0; j < 4 - 1; j++) {
out = ggml_concat(ctx->ggml_ctx, out, out_, 2);
}
}
}
}
if (wan2_2) {
Expand Down
Loading