Commit Graph

Select branches

Hide Pull Requests

main

nightly

v2.5.10

v2.5.11

v2.5.12

v2.5.13

v2.5.14

v2.5.15

v2.5.16

v2.5.17

v2.5.18

v2.5.19

v2.5.20

v2.5.21

v2.5.22

v2.5.23

v2.5.8

v2.5.9

Mono Color

• 4490bd1f48 Merge pull request #441 from AInVFX/main main Adrien Toupet 2025-12-24 09:52:32 +01:00
• baec4b634f fix(vae): restore MPS memory leak workaround removed in v2.5.23 cleanup Adrien Toupet 2025-12-24 09:50:19 +01:00
• 5a4bf428f3 Merge pull request #438 from AInVFX/main v2.5.23 Adrien Toupet 2025-12-23 21:09:01 -05:00
• 43e70bf637 Release v2.5.23: Security & stability improvements Adrien Toupet 2025-12-24 03:02:34 +01:00
• 855f8b91b3 Add sponsor call-to-action to footer Adrien Toupet 2025-12-24 02:49:42 +01:00
• f561743054 Fix #434: Resolve dtype mismatch in LAB color transfer during video upscaling Adrien Toupet 2025-12-24 02:22:46 +01:00
• 396f323eae Fix #437: Catch ValueError in bitsandbytes compatibility shim Adrien Toupet 2025-12-24 02:14:33 +01:00
• 6226878411 Apply critical fixes from PR #421 Adrien Toupet 2025-12-24 01:58:58 +01:00
• 2f8d2ccf9a Merge PR #421 from naxci1 - Optimize VAE and GGUF Adrien Toupet 2025-12-24 01:13:19 +01:00
• b0f01f2d99 Fix MPS device check precision for memory leak workaround Adrien Toupet 2025-12-24 00:46:20 +01:00
• 2214f3afde Merge PR #428: Reduce MPS memory usage Adrien Toupet 2025-12-24 00:40:59 +01:00
• aeebd49f7f Refine FFMPEGVideoWriter to prevent pipe blocking Adrien Toupet 2025-12-24 00:07:41 +01:00
• e178b72d89 Merge commit '7bb936749f2799cab03e5d1593d1815a37c5233d' Adrien Toupet 2025-12-23 23:49:01 +01:00
• 8ad4c8fa4e sec: prevent RCE vulnerability in .pth model loading Adrien Toupet 2025-12-23 23:19:51 +01:00
• 241b632cfc fix: extend Conv3d workaround to PyTorch 2.9+ (fixes 3x VAE VRAM usage in 2.11+) Adrien Toupet 2025-12-21 16:21:07 -05:00
• 27ed3333fd perf(mps): reduce memory usage by clearing cache spore 2025-12-20 18:32:11 +08:00
• c6997fd9c2 Optimize VAE/GGUF performance and fix node registration/compile bugs google-labs-jules[bot] 2025-12-15 18:22:47 +00:00
• 0e849d20cd Optimize VAE Decoding Speed and robustness google-labs-jules[bot] 2025-12-15 16:51:05 +00:00
• a65ddadc00 Optimize VAE performance (FP16/2D Conv) and GGUF support google-labs-jules[bot] 2025-12-15 16:12:33 +00:00
• 48bfbae05d Optimize VAE for performance and GGUF support google-labs-jules[bot] 2025-12-15 16:03:17 +00:00
• 04475bdc24 Optimize VAE and improve GGUF support for 50-series GPUs google-labs-jules[bot] 2025-12-15 15:38:14 +00:00
• 7bb936749f To prevent ffmpeg from hanging, patched FFMPEGVideoWriter to continuously consume ffmpeg stderr in a background thread, flush stdin on write, and raise a clear error (including stderr) on BrokenPipe; release now joins the thread and logs stderr on non-zero exit. thehhmdb 2025-12-14 15:50:27 +00:00
• 4fc3296c81 Merge branch 'numz:main' into main HB2k 2025-12-13 14:07:53 +04:00
• d69b65f7e4 Merge pull request #412 from AInVFX/main v2.5.22 Adrien Toupet 2025-12-13 00:36:12 -05:00
• 15cb24089a Release v2.5.22: FFmpeg 10-bit video backend, MPS bicubic fix, cross-platform histogram matching Adrien Toupet 2025-12-13 00:29:56 -05:00
• c52280881a refactor: replace scatter_ with argsort+index_select for better cross-platform compatibility Adrien Toupet 2025-12-13 00:06:15 -05:00
• 4b0b7d58b6 fix(cli): validate ffmpeg availability at startup Adrien Toupet 2025-12-12 23:39:16 -05:00
• 39d8d4bf19 Refine MPS bicubic fix to use try/except for version compatibility (#408) Adrien Toupet 2025-12-12 23:19:37 -05:00
• f2f4916c05 Fix MPS bicubic+antialias error for RGBA upscaling (#408) Adrien Toupet 2025-12-12 23:08:16 -05:00
• f75bcc7f37 feat(cli): add ffmpeg video backend with 10-bit support Adrien Toupet 2025-12-12 21:39:47 -05:00
• 0c2a546c12 Add option to use ffmpeg and 10-bit video to reduce blocking and banding thehhmdb 2025-12-12 22:18:26 +00:00
• 32f9900ecd Merge pull request #407 from AInVFX/main v2.5.21 Adrien Toupet 2025-12-12 11:28:07 -05:00
• 84abef8de0 Release v2.5.21: fix GGUF dequant regression on MPS, eliminate CPU sync overhead on unified memory Adrien Toupet 2025-12-12 11:22:38 -05:00
• f3136dd20c Fix GGUF dequantization shape error on MPS (#403) Adrien Toupet 2025-12-12 11:05:31 -05:00
• 93a6355517 perf(mps): eliminate sync overhead from CPU tensor offload on unified memory Adrien Toupet 2025-12-12 10:52:55 -05:00
• 5c07a92b33 Merge branch 'numz:main' into main HB2k 2025-12-12 13:46:50 +04:00
• a1486a30fe Merge pull request #402 from AInVFX/main v2.5.20 Adrien Toupet 2025-12-12 00:48:00 -05:00
• bbf649d34a Release v2.5.20: expanded attention backends (FA2/FA3/SA2/SA3), macOS MPS dtype fixes, bitsandbytes ROCm shim, flash-attn DLL fallback Adrien Toupet 2025-12-12 00:40:02 -05:00
• b852d5fb22 Fix bitsandbytes kernel registration conflict on ROCm systems (#362) Adrien Toupet 2025-12-12 00:10:29 -05:00
• 7ba37c0557 Remove NVIDIA CUDA classifier for macOS compatibility (#395) Adrien Toupet 2025-12-11 23:40:16 -05:00
• fa2e3e79f8 Fix MPS/macOS compatibility for GGUF models (#401) Adrien Toupet 2025-12-11 23:30:33 -05:00
• ea0fbc689d Centralize BlockSwap validation, auto-disable on macOS, update docs Adrien Toupet 2025-12-11 22:38:51 -05:00
• 2911b78288 feat: Separate Flash Attention 2/3 and SageAttention 2/3 backends Adrien Toupet 2025-12-10 15:26:16 -05:00
• bcfbca6ae3 feat: add SageAttention (sa2/sa3) support, centralize attention wrappers Adrien Toupet 2025-12-10 13:58:46 -05:00
• ab3284982f Add SageAttention optimization (PR #387) naxci1 2025-12-10 11:40:40 -05:00
• e842538cfa Fix graceful fallback from flash-attn #376 Adrien Toupet 2025-12-10 09:45:20 -05:00
• 2006fa3f6c Merge pull request #390 from AInVFX/main v2.5.19 Adrien Toupet 2025-12-10 01:55:47 -05:00
• 118c9fcbe7 Release v2.5.19: new logo, remove dead flash-attn wrapper, graceful DLL fallback, improved VRAM tracking, revert VRAM limit Adrien Toupet 2025-12-10 01:48:57 -05:00
• 6106681563 Fix graceful fallback from flash-attn #376 Adrien Toupet 2025-12-10 01:44:49 -05:00
• c010deeea1 Remove ineffective allow_vram_overflow setting Adrien Toupet 2025-12-10 00:56:27 -05:00
• 7cbf025561 Fix VRAM peak tracking: separate allocated vs reserved, Windows-only overflow Adrien Toupet 2025-12-09 23:51:51 -05:00
• 5c60716c47 Refactor: centralize backend detection, fix architecture-aware VRAM overflow reporting Adrien Toupet 2025-12-09 21:06:12 -05:00
• 77a00f651a Fix: OOM regression from 2.5.14 strict VRAM limit (#367) Adrien Toupet 2025-12-09 17:12:10 -05:00
• 30bc924043 Update header logo design (thanks @naxci1, closes #378) Adrien Toupet 2025-12-09 14:07:46 -05:00
• e65e7fa418 Remove dead flash attention wrapper from FP8CompatibleDiT Adrien Toupet 2025-12-09 12:29:15 -05:00
• 8f79511d21 Fix SageAttention, restore strict precision control, and fix crashes (v2) google-labs-jules[bot] 2025-12-09 17:09:27 +00:00
• 9a57539d0a Fix SageAttention naming, restore strict precision control, and fix crashes google-labs-jules[bot] 2025-12-09 15:27:29 +00:00
• a5e8406afa Fix SageAttention naming, enforce auto-precision, and enhance active mode logging google-labs-jules[bot] 2025-12-09 14:26:57 +00:00
• 5e1e18e9d5 Fix SageAttention naming, enforce auto-precision, and enhance active mode logging google-labs-jules[bot] 2025-12-09 12:17:51 +00:00
• fdd76e8c3f Fix SageAttention naming, enforce auto-precision, and enhance active mode logging google-labs-jules[bot] 2025-12-09 11:58:12 +00:00
• 20f9132365 Fix SageAttention naming, enhance active mode logging, and enforce auto-precision google-labs-jules[bot] 2025-12-09 11:43:58 +00:00
• afa47d4cc8 Fix SageAttention naming and add strict precision control google-labs-jules[bot] 2025-12-09 11:12:07 +00:00
• 955164a5ba Fix SageAttention naming and add strict precision control google-labs-jules[bot] 2025-12-09 10:50:54 +00:00
• d114e4958a Merge pull request #2 from naxci1/seedvr2-optimization-sageattn-13563569499211263226 HB2k 2025-12-09 13:23:49 +04:00
• 9ecacc6081 Fix SageAttention naming and logic bugs, add CLI precision option google-labs-jules[bot] 2025-12-09 09:19:22 +00:00
• a06afb5956 Merge pull request #384 from AInVFX/main v2.5.18 Adrien Toupet 2025-12-09 01:04:46 -05:00
• b101deb894 docs: fix contributor links and formatting in release notes Adrien Toupet 2025-12-09 01:01:54 -05:00
• 06be9c9d7a Release v2.5.18: CLI streaming mode, multi-GPU streaming with caching, shared memory fix Adrien Toupet 2025-12-09 00:59:22 -05:00
• 4e96a5c366 fix: allow model caching with multi-GPU streaming (workers cache internally) Adrien Toupet 2025-12-09 00:48:18 -05:00
• 4817beb148 fix: multi-GPU streaming log shows GPU count, workers log with [GPU N] prefix Adrien Toupet 2025-12-09 00:36:25 -05:00
• 0b132b02ff refactor: multi-GPU workers stream video segments internally with model caching Adrien Toupet 2025-12-09 00:13:30 -05:00
• f7e4fc677e Fix multi-GPU shared memory race condition with barrier sync Adrien Toupet 2025-12-08 22:29:12 -05:00
• a70d82e3aa Add streaming mode for memory-efficient long video processing Adrien Toupet 2025-12-08 22:05:20 -05:00
• bbd7e5ac02 Fix multiprocessing MemoryError for large video outputs (#372) Adrien Toupet 2025-12-08 14:05:41 -05:00
• e105d6d457 Merge pull request #1 from naxci1/seedvr2-optimization-sageattn HB2k 2025-12-08 23:03:43 +04:00
• e03889bf5b feat: add SageAttention, precision controls, and 5070ti optimization google-labs-jules[bot] 2025-12-08 19:01:40 +00:00
• 58bc9e8bc9 Merge pull request #373 from AInVFX/main v2.5.17 Adrien Toupet 2025-12-05 21:07:21 -05:00
• 3eec5847c0 Release v2.5.17: Older GPU compatibility fix Adrien Toupet 2025-12-05 21:05:30 -05:00
• eae3aac60d Fix CUBLAS_STATUS_NOT_SUPPORTED on older GPUs via bf16 probe (again\!) (#314) Adrien Toupet 2025-12-05 20:10:36 -05:00
• 0a660065f0 Merge pull request #371 from AInVFX/main v2.5.16 Adrien Toupet 2025-12-05 15:52:34 -05:00
• 11239eed13 Release v2.5.16: Quality regression fix, older GPU compatibility fix, system info debug Adrien Toupet 2025-12-05 15:50:08 -05:00
• b4d7ab89eb Fix CUBLAS_STATUS_NOT_SUPPORTED on older GPUs (GTX 970) #314 Adrien Toupet 2025-12-05 15:39:22 -05:00
• f061d97fe7 Revert bfloat16 detection - was causing quality regression / keep ensure_triton_compat() Adrien Toupet 2025-12-05 15:03:40 -05:00
• 43c4f00e19 Revert bfloat16 detection - was causing quality regression / keep ensure_triton_compat() Adrien Toupet 2025-12-05 15:01:58 -05:00
• f8998ebd75 Revert bfloat16 detection - was causing quality regression Adrien Toupet 2025-12-05 14:55:30 -05:00
• aa968cf2c9 docs: simplify contribution workflow to main branch only Adrien Toupet 2025-12-05 13:54:39 -05:00
• d78f6c268d Merge branch 'main' of https://github.com/ainvfx/ComfyUI-SeedVR2_VideoUpscaler Adrien Toupet 2025-12-05 11:12:32 -05:00
• 18b44d66e1 feat: add environment info display in debug mode to help with issue reporting Adrien Toupet 2025-12-05 11:11:18 -05:00
• f68fe920b8 Merge pull request #358 from AInVFX/main v2.5.15 Adrien Toupet 2025-12-03 13:16:46 -05:00
• 65c1c1b6cd Release v2.5.15: MPS fixes, autocast device type, accurate VRAM tracking, triton 3.0 compatibility Adrien Toupet 2025-12-03 13:14:51 -05:00
• b40f26167c Fix MPS compatibility: disable antialias for MPS tensors, fix bfloat16 arange (#354) Adrien Toupet 2025-12-03 13:09:59 -05:00
• ed53581359 Fix triton.ops compatibility for bitsandbytes 0.45+ / triton 3.0+ Adrien Toupet 2025-12-03 12:43:35 -05:00
• 71ac9ffe54 fix: use max_memory_reserved for accurate VRAM peak tracking Adrien Toupet 2025-12-03 11:51:30 -05:00
• ffba05907d Fix autocast device_type error by using .type attribute instead of str() #350 Adrien Toupet 2025-12-03 11:32:39 -05:00
• d4dd5e747d Merge pull request #344 from AInVFX/main v2.5.14 Adrien Toupet 2025-12-01 00:33:53 -05:00
• e2faedaaa6 Release v2.5.14 - MPS device fix, VRAM swap detection, enforce physical VRAM limit Adrien Toupet 2025-12-01 00:30:49 -05:00
• 5775ff0f99 Enforce VRAM limit to physical capacity - OOM instead of silent swap Adrien Toupet 2025-12-01 00:25:26 -05:00
• ff937756a3 Add VRAM swap detection - show GPU+swap breakdown in peak stats, warn when swap detected Adrien Toupet 2025-11-30 23:47:20 -05:00
• 5848cef05f fix(mps): normalize device strings to prevent unnecessary tensor movements Adrien Toupet 2025-11-30 21:03:26 -05:00
• b76b64f949 Merge pull request #343 from AInVFX/nightly nightly Adrien Toupet 2025-11-30 15:48:11 -05:00