nodes/hao-ai-lab-FastVideo
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

[feat] v2: register Wan2.1-I2V-14B 480P/720P (reuse the GPU-verified i2v card)

Browse Source 
The 14B i2v variants reuse the Wan2.1 i2v card/path proven on Fun-1.3B-InP — just per-variant params
(480P flow_shift 3.0 / 480x832, 720P flow_shift 5.0 / 720x1280). Registry resolution + caps verified;
specific 14B weights GPU-pending (same generic Wan i2v loader path that Fun-InP validated). i2v cluster
now supported; roadmap updated (11 models ported).
This commit is contained in:
Will Lin
2026-07-04 17:15:34 +00:00
committed by Peiyuan Zhang
parent 4a56274d10
commit 7464ef8308
2 changed files with 12 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files
examples/inference/basic/V2_PORTING_STATUS.md
+4 -1
View File
@@ -55,7 +55,10 @@ Wan2.1 model — covered by wan21; those runtime modes are out of scope.
| `basic_ltx2_3_distilled_i2v*.py` | LTX-2.3-Distilled I2V | Image conditioning on the ltx2 single-stage program (encode the input image + condition the denoise). |
## Backlog — goal: port ALL fastvideo models (63 total: 8 ported · 21 reuse-arch · 34 new-arch)
Ported (8): the 7 above + **Wan2.1-T2V-14B** (bucket B — reuses the Wan recipe/adapter; 720p defaults).
Ported (11): the 7 above + **Wan2.1-T2V-14B** + the **Wan2.1 i2v cluster** — Wan2.1-Fun-1.3B-InP
(GPU-verified: CLIP encoder + first-frame [mask|cond] -> 36ch DiT) and Wan2.1-I2V-14B-480P/720P (reuse the
same i2v card; 14B weights GPU-pending). i2v required a real sub-system: the image-encoder adapter, the
loop's i2v conditioning hooks, and the Wan adapter's 36ch concat (all landed + T2V-unregressed).
⚠️ Not as trivial as the bucket split implies: the `FastVideo/*` distilled checkpoints (FastWan DMD, VSA,
Turbo) carry **gated-attention params** (`to_gate_compress`) the generic `WanTransformer3DModel` loader
v2/registry.py
+8 -1
View File
@@ -55,9 +55,16 @@ def _entries() -> list[ModelEntry]:
return [
ModelEntry(("Wan-AI/Wan2.1-T2V-1.3B-Diffusers",), build_wan21_card, build_wan_t2v_program),
ModelEntry(("Wan-AI/Wan2.1-T2V-14B-Diffusers",), build_wan_t2v_14b_card, build_wan_t2v_program),
# Wan2.1 i2v (Fun-InP): CLIP image encoder + first-frame VAE conditioning ([mask|cond] -> 36ch DiT).
# Wan2.1 i2v cluster: CLIP image encoder + first-frame VAE conditioning ([mask|cond] -> 36ch DiT).
# Fun-1.3B-InP is GPU-verified; the 14B variants reuse the same i2v card/path (weights GPU-pending).
ModelEntry(("weizhou03/Wan2.1-Fun-1.3B-InP-Diffusers",),
build_wan21_i2v_card, build_wan21_i2v_program),
ModelEntry(("Wan-AI/Wan2.1-I2V-14B-480P-Diffusers",),
lambda: build_wan21_i2v_card("wan2.1-i2v-14b-480p", flow_shift=3.0, height=480, width=832),
build_wan21_i2v_program),
ModelEntry(("Wan-AI/Wan2.1-I2V-14B-720P-Diffusers",),
lambda: build_wan21_i2v_card("wan2.1-i2v-14b-720p", flow_shift=5.0, height=720, width=1280),
build_wan21_i2v_program),
ModelEntry(("Wan-AI/Wan2.2-TI2V-5B-Diffusers",), build_wan22_ti2v_card, build_wan_t2v_program),
ModelEntry(("Wan-AI/Wan2.2-T2V-A14B-Diffusers",), build_wan22_a14b_card, build_wan_t2v_program),
ModelEntry(("wlsaidhi/SFWan2.1-T2V-1.3B-Diffusers",), build_wan_causal_card, build_wan_causal_program),