Commit Graph

Select branches

Hide Pull Requests

docs/attribution-scope-20260709

main

Mono Color

• 95d909a787 docs: update README and NOTICE for official Animate2 Embeds + SCAIL_2 Flow Sampler integration (no separate Animate2 Sampler) main wuwukasi 2026-09-14 22:46:54 +08:00
• ef2139524c add: official-compatible Animate2 flow nodes wuwukasi 2026-09-13 23:58:43 +08:00
• 4327a9fced fix: align wananimate loop prefix preview latents wuwukasi 2026-07-28 23:37:50 +08:00
• 155f2cf7f7 fix: restore scail2 flow logging wuwukasi 2026-07-28 23:21:15 +08:00
• 4d58428e60 fix: optimize scail2 flow deferred encoding wuwukasi 2026-07-28 23:03:28 +08:00
• dd40829210 fix: show flow sampler progress previews wuwukasi 2026-07-28 21:09:47 +08:00
• bccd050c1d fix: confine multitalk output paths wuwukasi 2026-07-27 00:41:56 +08:00
• 24fad36766 fix: relax SCAIL-2 flow sampler VAE input wuwukasi 2026-07-26 19:13:33 +08:00
• 20bb48dc46 add: official-compatible SCAIL-2 flow nodes wuwukasi 2026-07-26 18:29:26 +08:00
• bd01021a52 add: support scail2 auto drift colormatch wuwukasi 2026-07-25 22:19:41 +08:00
• 816345fb93 docs: recommend easy sampler nodes wuwukasi 2026-07-24 19:26:23 +08:00
• 5e16d6a746 add: add easy sampler settings node wuwukasi 2026-07-24 19:19:34 +08:00
• a5e4605b23 add: add WanAnimatePlus easy sampler wuwukasi 2026-07-24 18:05:51 +08:00
• c22bcc2ff6 fix: restore direct SCAIL-2 colormatch wuwukasi 2026-07-24 16:55:31 +08:00
• 15ec0f2beb fix: skip colormatch shared memory when shm is small wuwukasi 2026-07-22 00:19:50 +08:00
• fabbf1d888 fix: cast mask blends for PyTorch bool tensors wuwukasi 2026-07-21 15:01:04 +08:00
• b36b9f9d04 docs: clarify RunningHub free workflow links wuwukasi 2026-07-20 15:10:44 +08:00
• 13bf1aea53 docs: add RunningHub workflow examples wuwukasi 2026-07-20 14:59:35 +08:00
• 7ee4202985 fix: handle comfy quant block swap apply wuwukasi 2026-07-19 20:14:07 +08:00
• 513be99ff6 fix: support ConvRot W4A4 native quant wuwukasi 2026-07-19 16:58:51 +08:00
• 0318a8c69a fix: use shared memory for SCAIL-2 colormatch wuwukasi 2026-07-19 13:40:00 +08:00
• 5fa6ed0cbc fix: cast SCAIL-2 direct freeze mask wuwukasi 2026-07-19 13:12:55 +08:00
• 326e5a7ccf fix: log SCAIL-2 loop chunk timings wuwukasi 2026-07-19 12:35:38 +08:00
• 2c8fe1c847 fix: normalize SCAIL-2 window handling wuwukasi 2026-07-18 23:18:41 +08:00
• 015f8700fa Normalize SCAIL-2 loop frame windows wuwukasi 2026-07-18 14:28:09 +08:00
• 0abb821cd4 add: SCAIL-2 example workflow wuwukasi 2026-07-18 20:15:59 +08:00
• 6275a340b4 Optimize SCAIL-2 context memory usage wuwukasi 2026-07-18 15:51:54 +08:00
• b8fd845137 Optimize SCAIL-2 two-phase freeze replacement docs/attribution-scope-20260709 wuwukasi 2026-07-13 20:40:20 +08:00
• 3a3752fb93 Optimize SCAIL-2 two-phase freeze replacement wuwukasi 2026-07-13 20:40:20 +08:00
• e8954be6fa Route comfy_quant FP8 through legacy fp8 path wuwukasi 2026-07-13 20:11:26 +08:00
• 1607e3cead Simplify README attribution notices wuwukasi 2026-07-13 16:57:54 +08:00
• d1911f64a2 Delete SCAIL2_LOOP_FAST_PATH_PLAN.md wuwukasi 2026-07-13 16:47:35 +08:00
• eef2f95109 Merge docs/attribution-scope-20260709 into main wuwukasi 2026-07-13 16:44:26 +08:00
• 0f9f4744d5 Add SCAIL-2 two-phase loop sampling settings wuwukasi 2026-07-13 14:50:59 +08:00
• d3809441e2 docs: clarify downstream notice rights wuwukasi 2026-07-12 19:39:40 +08:00
• d85c0966a1 chore: clarify RoPE and upstream notice scope wuwukasi 2026-07-12 02:44:02 +08:00
• 396bd4e868 Merge pull request #19 from wuwukaka/docs/attribution-scope-20260709 wuwukasi 2026-07-09 21:38:34 +08:00
• 932f7b82ab docs: clarify attribution scope wuwukasi 2026-07-09 16:25:54 +08:00
• bdfbf5b7c2 perf: enable Q×Q GEMM for fp8/nvfp4 comfy_quant formats wuwukasi 2026-07-09 16:00:04 +08:00
• 575fec5f68 fix: generalize comfy_quant block swap device resolver to all formats wuwukasi 2026-07-07 23:55:43 +08:00
• 6363386c57 style: translate Chinese comments to English in int8_tensorwise support wuwukasi 2026-07-07 23:37:10 +08:00
• f598ef7ebb feat: add int8_tensorwise (convrot) support to comfy_quant_linear wuwukasi 2026-07-07 23:32:38 +08:00
• 96aa2b8ca9 Isolate SCAIL2 colormatch worker wuwukasi 2026-07-07 22:24:01 +08:00
• c9434dcade Fix WanAnimate context control alignment wuwukasi 2026-07-07 20:48:18 +08:00
• 028b4413fb Add WanAnimate single-frame prefix encoding wuwukasi 2026-07-05 02:04:08 +08:00
• f1f1f0cff4 Adjust SCAIL-2 bg image ordering wuwukasi 2026-07-03 22:02:21 +08:00
• 8af48ddafe Clean stale SCAIL-2 loop caches wuwukasi 2026-07-03 21:07:34 +08:00
• 02c3d10907 Optimize SCAIL-2 loop caching and attention dispatch wuwukasi 2026-07-03 15:18:30 +08:00
• d51d7cb477 Handle stale quant metadata for LoRA wuwukasi 2026-07-02 16:52:30 +08:00
• 338ac6ae69 Optimize SCAIL-2 loop memory usage wuwukasi 2026-06-30 01:50:06 +08:00
• ae04a76290 Fix SCAIL2 loop colormatch reference handling wuwukasi 2026-06-27 18:31:17 +08:00
• 78f9e8531e Fix NVFP4 block swap materialization wuwukasi 2026-06-27 18:12:47 +08:00
• 8372807587 fix: reject native quant lora fallback wuwukasi 2026-06-27 14:28:57 +08:00
• 4cd48c0c9d fix: reduce native quant lora activation memory wuwukasi 2026-06-27 13:54:39 +08:00
• 94e93f8b08 chore: hide scail2 timing controls wuwukasi 2026-06-27 13:25:47 +08:00
• 8caba2c5cb fix: refine scail2 loop colormatch reference wuwukasi 2026-06-27 13:12:59 +08:00
• dfc9489852 fix: color match scail2 loop output frames wuwukasi 2026-06-27 12:16:56 +08:00
• 8e36a5c1d9 fix: reduce nvfp4 lora memory usage wuwukasi 2026-06-27 11:46:24 +08:00
• 8c6201ed8d fix: handle scale-only nvfp4 packed weights wuwukasi 2026-06-27 11:25:18 +08:00
• 701271ccfd fix: pass raw time embedding to uni3c fast path wuwukasi 2026-06-22 22:47:01 +08:00
• bf30669318 fix: rematerialize comfy quant weights after offload wuwukasi 2026-06-22 19:04:06 +08:00
• 090cd6b692 fix: support uni3c in scail2 fast path wuwukasi 2026-06-22 18:16:36 +08:00
• 9fcec0bb8b fix: pass uni3c through scail2 loop chunks wuwukasi 2026-06-22 17:48:28 +08:00
• 974d76c27e fix: preserve nvfp4 scale metadata aliases wuwukasi 2026-06-18 22:43:31 +08:00
• 64ed5fd29a docs: clarify attribution scope wuwukasi 2026-06-18 22:22:43 +08:00
• 759e002478 docs: require video attribution wuwukasi 2026-06-18 22:14:33 +08:00
• 9f897e368f fix: infer nvfp4 metadata for packed lora targets wuwukasi 2026-06-18 16:14:34 +08:00
• de32103a6d fix: match native quant lora metadata by shape wuwukasi 2026-06-18 15:53:46 +08:00
• 06d769a3d9 fix: rebind native quant lora metadata in skyreels wuwukasi 2026-06-18 15:44:31 +08:00
• 455feec14b fix: bind native quant lora targets by prefix wuwukasi 2026-06-18 15:42:07 +08:00
• deb8409a4b fix: rebind native quant metadata before lora wuwukasi 2026-06-18 15:20:51 +08:00
• 435870347b fix: bind native quant metadata for lora wuwukasi 2026-06-18 14:41:57 +08:00
• 2884562d58 fix: handle native quant packed linear weights wuwukasi 2026-06-18 13:55:44 +08:00
• 422ca1ee07 fix: use logical shapes for native quant weights wuwukasi 2026-06-18 12:40:19 +08:00
• 717a9dff5c fix: dequantize native quant weights before LoRA wuwukasi 2026-06-18 12:27:26 +08:00
• c2d84d68f1 fix: keep quant scale weights on linear weights wuwukasi 2026-06-18 12:05:19 +08:00
• e87ae8c6b2 fix: chunk SCAIL2 mask resizing wuwukasi 2026-06-18 11:45:58 +08:00
• 8868eedff5 Add Comfy native quantized linear loading wuwukasi 2026-06-17 22:44:59 +08:00
• 8d4df80f18 Add SCAIL2 loop fast path wuwukasi 2026-06-17 22:27:01 +08:00
• 1cec8a55e7 fix: export clip vision encode v2 from package init wuwukasi 2026-06-15 21:23:16 +08:00
• a578135916 feat: add clip vision encode v2 for multi-image batches wuwukasi 2026-06-15 20:56:20 +08:00
• 22555324f4 Sync SCAIL-2 replacement pose RoPE and prefix mask background wuwukasi 2026-06-14 21:23:45 +08:00
• 9d7cb804d9 revert: restore nodes.py to afc6c45 state wuwukasi 2026-06-14 17:20:30 +08:00
• 167b4308bd revert: restore compatibility by returning files to 76173c0 state wuwukasi 2026-06-14 17:09:01 +08:00
• 75f24e7722 fix: restore direct model pipeline access for sampler flags wuwukasi 2026-06-14 16:28:44 +08:00
• e055cd9087 fix: avoid dict-style get on WanVideoModel pipeline wuwukasi 2026-06-14 16:26:08 +08:00
• 38ef36fbca fix: preserve legacy fp8 fast compatibility flag wuwukasi 2026-06-14 15:19:03 +08:00
• 9748bd56ef fix: read runtime flags from patcher model wuwukasi 2026-06-14 15:14:33 +08:00
• aab12f0d93 refactor: separate regular FP8 and MXFP8 runtime flags wuwukasi 2026-06-14 15:04:58 +08:00
• 09f52aa0b9 fix: allow MXFP8 fallback with unmerged LoRAs wuwukasi 2026-06-14 14:54:54 +08:00
• b9ffae6131 fix: stabilize MXFP8 fast path fallback behavior wuwukasi 2026-06-14 14:42:31 +08:00
• 2fce828d10 fix: align MXFP8 scale tensors with weight devices wuwukasi 2026-06-14 14:12:59 +08:00
• 85286e02c6 fix: allow MXFP8 unmerged LoRA sampler fallback wuwukasi 2026-06-14 13:57:30 +08:00
• 439c19e76c fix: keep MXFP8 usable without forced BF16 fallback wuwukasi 2026-06-14 13:38:23 +08:00
• 8681c43b8a fix: reduce MXFP8 forward memory usage wuwukasi 2026-06-14 12:34:40 +08:00
• 76d8f354a6 fix: restore original LoRA memory behavior wuwukasi 2026-06-14 12:21:10 +08:00
• 9551d8e810 fix: use comfy patch weights in MXFP8 LoRA fallback wuwukasi 2026-06-14 01:27:18 +08:00
• 092ed56f8a fix: avoid custom_linear import cycle wuwukasi 2026-06-14 01:09:11 +08:00
• 66f96b0e23 chore: ignore claude workspace files wuwukasi 2026-06-14 00:55:31 +08:00
• 780d4a3502 perf: improve MXFP8 unmerged LoRA support wuwukasi 2026-06-14 00:54:53 +08:00