Commit Graph

7198 Commits

Author SHA1 Message Date
Disty0 9e52d0c1fb SDNQ add SVDQuant quantization method 2025-10-05 22:50:30 +03:00
Vladimir Mandic 268798a24e add framepack granular job tracking
Signed-off-by: Vladimir Mandic <mandic00@live.com>
2025-10-05 10:39:55 -04:00
Disty0 428600613a SDNQ fix new transformers again 2025-10-05 15:30:15 +03:00
Disty0 a164f3e0c2 SDNQ Improve UINT3 and below quant speed 2025-10-05 03:12:05 +03:00
Vladimir Mandic 8325e886c7 add typo to legacy compatibility options
Signed-off-by: Vladimir Mandic <mandic00@live.com>
2025-10-04 18:11:52 -04:00
Vladimir Mandic c530167cbe qwen multi-image edits
Signed-off-by: Vladimir Mandic <mandic00@live.com>
2025-10-04 18:06:50 -04:00
Vladimir Mandic 8b698ed67f upadte qwen pruning and allow hf models in subfolders
Signed-off-by: Vladimir Mandic <mandic00@live.com>
2025-10-04 15:49:20 -04:00
Disty0 f2e12a682f SDNQ remove use_contiguous_mm path in re_quant 2025-10-04 19:17:05 +03:00
Disty0 df142afe81 don't use triton mm for nvidia 2025-10-04 18:48:03 +03:00
Disty0 5c5d7d5a86 cleanup 2025-10-04 18:38:18 +03:00
Disty0 99113947bf SDNQ add RDNA2 INT8 support via Triton 2025-10-04 18:31:25 +03:00
Vladimir Mandic 54ae18a611 update nunchaku
Signed-off-by: Vladimir Mandic <mandic00@live.com>
2025-10-03 12:23:24 -04:00
Disty0 95a7da7e75 SDNQ use non-contiguous re-quantize 2025-10-03 18:54:58 +03:00
Vladimir Mandic a6108dd6df add qwen pruning variants
Signed-off-by: Vladimir Mandic <mandic00@live.com>
2025-10-03 11:36:14 -04:00
Disty0 54acf1760b Make SDNQ scales compatible with balanced offload 2025-10-03 18:13:55 +03:00
Disty0 c5cab96223 SDNQ simplify check_mats 2025-10-03 02:58:17 +03:00
Vladimir Mandic 7325f9dbae dont import guiders until needed
Signed-off-by: Vladimir Mandic <mandic00@live.com>
2025-10-02 16:28:34 -04:00
Disty0 34c2a624aa SDNQ autodetect fp8 tw fallback and disable dynamic compile 2025-10-02 19:40:07 +03:00
Vladimir Mandic f245506bf2 cleanup hf login
Signed-off-by: Vladimir Mandic <mandic00@live.com>
2025-10-02 08:17:35 -04:00
CalamitousFelicitousness 78711fb1d4 Merge branch 'dev' into patch-2 2025-10-01 20:58:58 +01:00
CalamitousFelicitousness 78820a14dc Allow VLM temp setting temperature to 0
Allow VLM temp setting temperature to 0
2025-10-01 20:52:04 +01:00
Disty0 81bb2b99ef update florence promptgen repo ids 2025-10-01 21:43:02 +03:00
Vladimir Mandic 22074f4727 cleanup vqa
Signed-off-by: Vladimir Mandic <mandic00@live.com>
2025-10-01 12:02:55 -04:00
Vladimir Mandic 5d0a3e5e8a fix microsoft-florence
Signed-off-by: Vladimir Mandic <mandic00@live.com>
2025-10-01 10:58:52 -04:00
Disty0 03382bdd4c SDNQ simplify check_mats 2025-10-01 01:35:51 +03:00
Vladimir Mandic 8c2b76a8da add buffers and cache info
Signed-off-by: Vladimir Mandic <mandic00@live.com>
2025-09-30 14:36:55 -04:00
Vladimir Mandic 201ca28300 improve ram monitor
Signed-off-by: Vladimir Mandic <mandic00@live.com>
2025-09-30 14:27:43 -04:00
Disty0 0c1d34721c SDNQ use contiguous for intel 2025-09-30 02:37:58 +03:00
Vladimir Mandic 2621aecacc interrrupt will now show last known preview image
Signed-off-by: Vladimir Mandic <mandic00@live.com>
2025-09-29 19:25:50 -04:00
Disty0 6b67a9d0c4 SDNQ add check_mats to matmul 2025-09-30 01:58:13 +03:00
Vladimir Mandic 93bdca0670 fix image str decode and update requirements
Signed-off-by: Vladimir Mandic <mandic00@live.com>
2025-09-29 10:56:03 -04:00
Vladimir Mandic 5d1bfc2bfc cleanup
Signed-off-by: Vladimir Mandic <mandic00@live.com>
2025-09-28 21:36:16 -04:00
Vladimir Mandic c03a6928e4 reference models add size and release date info
Signed-off-by: Vladimir Mandic <mandic00@live.com>
2025-09-28 21:19:36 -04:00
Vladimir Mandic 70a2c209b1 cleanup
Signed-off-by: Vladimir Mandic <mandic00@live.com>
2025-09-28 20:22:25 -04:00
Vladimir Mandic 8cc9c23f98 add directml and rocm note to changelog
Signed-off-by: Vladimir Mandic <mandic00@live.com>
2025-09-28 18:44:42 -04:00
Vladimir Mandic c661497b87 linting
Signed-off-by: Vladimir Mandic <mandic00@live.com>
2025-09-28 18:09:09 -04:00
Vladimir Mandic eb025eaf31 remove split-attention and add attention slicing option
Signed-off-by: Vladimir Mandic <mandic00@live.com>
2025-09-28 18:01:34 -04:00
Disty0 1b45c145e9 SDNQ re-enable dynamic compile 2025-09-28 20:40:24 +03:00
Disty0 6f2145fa57 revert offload_buffers 2025-09-28 16:47:31 +03:00
Vladimir Mandic 749878289f document vae changes
Signed-off-by: Vladimir Mandic <mandic00@live.com>
2025-09-28 09:25:38 -04:00
Vladimir Mandic 9e6fdaa5cc vae slicing and tiling enabled on low and med vram, off otherwise
Signed-off-by: Vladimir Mandic <mandic00@live.com>
2025-09-28 09:24:07 -04:00
Disty0 02f03d56a7 set offload_buffers to true on balanced offload 2025-09-28 16:23:29 +03:00
Disty0 4a70e82b0c ROCm always use numpy on cholesky 2025-09-28 14:05:20 +03:00
Disty0 a47959b114 move ROCm Windows hijacks outside of torch install 2025-09-28 13:33:15 +03:00
Seunghoon Lee 06b14b070b hijack torch.linalg.cholesky() 2025-09-28 18:43:37 +09:00
Seunghoon Lee 9c19e8a9b8 return correct item 2025-09-28 18:34:10 +09:00
Seunghoon Lee 6e8abf8dc3 hijack cholesky decomposition for therock pytorch 2025-09-28 18:03:02 +09:00
Seunghoon Lee 95ee0ed29c load amdhip64_6.dll instead of amdhip64.dll 2025-09-28 17:15:12 +09:00
Seunghoon Lee 0de3732884 prioritize python pacakage check over PATH 2025-09-28 11:31:43 +09:00
Disty0 587db51e9e Cleanup env vars and set TORCH_ROCM_AOTRITON_ENABLE_EXPERIMENTAL to 1 2025-09-28 00:22:00 +03:00