Spaces:
Running on Zero
Running on Zero
Commit History
-
Add ftfy (Wan V2V/VACE/I2V/Animate prompt_clean needs it) Running
Verified signatureb7bacf7
-
Wire V2V + FLF2V + TI2V-5B (batch 1) — prepare→generate, mirror-aware load
Verified signaturefa026b8 -
Enable Xet fast-transfer; revert default to Wan 2.1
Verified signaturebc37068 -
Prefetch retry on CDN drops
Verified signaturedac90ba -
MoE .to(cuda) on xlarge (bf16 fits 96GB)
Verified signatureefaa369 -
No download inside metered GPU call: startup prefetch + non-GPU prepare step
Verified signaturef128b1d -
I2V large/xlarge routing
Verified signature887dfbd -
Device-aware bf16 loading + T2V large/xlarge slice routing
Verified signature63d9d01
-
Load weights from HF CDN (fast) instead of the slow volume mount
Verified signature402bdc8 -
Stabilize: gate startup load off (restore RUNNING)
Verified signatureb0b97ce -
Module-level startup load + default Wan 2.1 14B + torchao 0.17 pin
Verified signature57fa516 -
Disable background warm (restart loop); lazy-load with adequate budget
Verified signaturee90d1e7 -
Post-launch background warm of Wan 2.1 T2V-14B (COW for workers)
Verified signatureadc308d -
Cold-load-aware ZeroGPU duration budgets
Verified signature073fc2e -
Stabilize startup: lazy model load + wan2.1-t2v-14b stitch bundle
Verified signatureLoad models lazily on first Generate instead of preloading the 27B MoE into main-process CPU RAM at import (which blocked demo.launch and wedged the Space in APP_STARTING). Adds models_meta/wan2.1-t2v-14b stitch bundle so single-transformer Wan 2.1 T2V-14B loads from the read-only mount. Deployed from branch wan-studio/phase1-stabilization.e2bcf63 -
README: proper OSS landing page + Apache 2.0 LICENSE
Verified signature529998e
-
Stitched-dir loader — zero downloads via mount + bundled JSONs
Verified signatureac419f5 -
Disable track_tqdm — UI scroll growth fix attempt
Verified signature03aa7a3 -
Preload T2V handle; split load from CUDA-attach
Verified signaturef973cdb -
Pass local snapshot path + worker probe
Verified signature6985fa8 -
Redirect HF cache to /tmp + visible snapshot_download
Verified signaturef9fb977 -
Trim preload to 1 model + LoRA (150GB cap)
Verified signaturea614d74 -
Bypass volume mount + add base model preloads
Verified signatured11d40a -
DEBUG — dump mounted-volume contents at startup
Verified signatureabd285a -
Fix ZeroGPU 422 — static size literal + WIP dev banner
Verified signatureaa0ee36 -
Wave K — mobile + tablet responsive (Linear theme preserved)
Verified signaturebf9c9e3
-
Phase 1 fix — drop MoE to large + CPU offload (PRO compat)
Verified signature0945b45 -
Wave I — port 7860 on HF Spaces (commit d92b33a)
Verified signatured7c6b90 -
Wave I — move theme/css to gr.Blocks() ctor (commit d7ab412)
Verified signaturec7e270d -
Wave I — import spaces first to fix CUDA-already-initialized (commit 6e2d426)
Verified signature2eeabfe -
Phase 1 — Wan Studio T2V/I2V on Wan 2.1 + 2.2 (commit fec10de)
Verified signature331f0f2
-
Bump gradio to 5.49 (fix HfFolder import error)
Verified signature458d612 -
Initial probe: detect ZeroGPU hardware (H200 vs Blackwell)
Verified signatureaebbb85 -
initial commit
Verified signaturebda7b26