Mario Limonciello (AMD)
85c88923b5
Switch over ROCm builds to artifacts both for stable and preview releases
...
AMD is in the process of moving over to a brand new build system (TheRock).
They have a preview release stream for it, and it includes some significant
performance uplifts.
To let people experiment with both stacks generate artifacts both for the
stable (7.2.1) and preview (7.12) stacks.
In both cases - ROCm must be installed first to use this artifact.
It's intentionally not bundled because so many ISAs are included and the
size of the artifact balloons to an untenable size.
2026-04-04 00:14:05 -05:00
leejet
87ecb95cbc
feat: add webp support ( #1384 )
2026-04-02 01:36:11 +08:00
Wagner Bruna and leejet
99c1de379b
feat: ancestral sampler implementations for flow models ( #1374 )
...
* feat: add support for the eta parameter to ancestral samplers
* feat: Euler Ancestral sampler implementation for flow models
* refine flow ancestral sampling and normalize eta defaults
---------
Co-authored-by: leejet <[email protected] >
2026-04-02 01:35:29 +08:00
leejet
09b12d5f6d
feat(cli): add metadata inspection mode ( #1381 )
2026-04-01 00:52:03 +08:00
leejet
6dfe945958
fix: use resolved image size in embedded metadata ( #1382 )
2026-03-31 23:55:49 +08:00
leejet
bf0216765a
feat: show tensor loading progress in MB/s or GB/s ( #1380 )
2026-03-31 23:06:44 +08:00
Wagner Bruna
4fe7a35939
feat(server): add generation metadata to png images ( #1217 )
2026-03-31 23:06:27 +08:00
Jan Ekström
4d5232083f
chore(server): link winsock2 for non-MSVC windows ( #1378 )
2026-03-31 22:10:34 +08:00
leejet
1d6cb0f8c3
refactor: split and simplify sample_k_diffusion samplers ( #1377 )
2026-03-31 00:32:14 +08:00
leejet
83e8f6f0af
refactor(server): split server endpoint registration ( #1376 )
2026-03-31 00:02:03 +08:00
leejet
8d878872d9
fix: disable extra T5 mask padding for Wan ( #1375 )
2026-03-30 23:10:07 +08:00
Wagner Bruna
02dd5e5dd2
refactor: move VAE tiling parameters to SDGenerationParams ( #1261 )
2026-03-30 23:09:38 +08:00
leejet
8f2967c006
sync: update ggml
2026-03-30 23:00:53 +08:00
leejet
f16a110f87
refactor: migrate generation pipeline to sd::Tensor ( #1373 )
2026-03-30 00:19:25 +08:00
stduhpf
ed88e215a2
refactor: simplify f8_e5m2_to_f16 function a little bit ( #1358 )
2026-03-30 00:14:33 +08:00
Wagner Bruna
6293ab5aaf
docs: update Spectrum info about DiT models ( #1360 )
2026-03-30 00:12:57 +08:00
leejet
545fac4f3f
refactor: simplify sample cache flow ( #1350 )
2026-03-17 00:28:03 +08:00
Tay
5265a5efa1
perf(z-image): switch to fused SwiGLU kernel ( #1302 )
2026-03-17 00:27:46 +08:00
leejet
84cbd88df1
style: remove redundant struct qualifiers for consistent C/C++ type usage ( #1349 )
2026-03-16 22:17:22 +08:00
Daniele
997bb11fb6
fix: correct encoder channels for flux2 ( #1346 )
2026-03-16 22:16:43 +08:00
leejet
862a6586cb
feat: add embedded WebUI ( #1207 )
2026-03-16 00:26:57 +08:00
leejet
61d8331ef3
ci: avoid cuda docker build timeout by using -j16
2026-03-15 18:39:29 +08:00
leejet
acc3bf1fdc
refactor: optimize the VAE architecture ( #1345 )
2026-03-15 16:57:42 +08:00
Kevin Nause
83eabd7c01
ci: add CUDA Dockerfile ( #1314 )
2026-03-15 16:46:01 +08:00
Wagner Bruna
630ee03f23
refactor: move all cache parameter defaults to the library ( #1327 )
2026-03-15 16:43:46 +08:00
Wagner Bruna
f6968bc589
chore: remove SD_FAST_SOFTMAX build flag ( #1338 )
2026-03-15 16:42:47 +08:00
rmatif
adfef62900
feat: add generic DiT support to spectrum cache ( #1336 )
2026-03-15 16:41:05 +08:00
JusteLeo
6fa7ca9317
docs: add Anima2 gguf download link to anima.md ( #1335 )
2026-03-15 16:40:14 +08:00
leejet
d6dd6d7b55
refactor: remove ununsed encode_video ( #1332 )
2026-03-10 00:36:09 +08:00
rmatif
dea4980f4e
feat: add spectrum caching method ( #1322 )
2026-03-10 00:35:32 +08:00
leejet
c8fb3d2458
fix: resolve SD1 Pix2Pix issue ( #1329 )
2026-03-08 00:28:05 +08:00
stduhpf
3d33caaef8
fix: make tiling work better when using circular ( #1299 )
2026-03-08 00:25:07 +08:00
WinkelCode
9b424db0f4
ci: change workflow owner of "actions-commit-hash" from "pr-mpt" to "prompt" ( #1323 )
2026-03-08 00:23:23 +08:00
rmatif
d95062737e
fix: ucache: normalize reuse error ( #1313 )
2026-03-04 23:50:45 +08:00
Korsar13
7c880f80c7
fix: avoid sd-server memory leak ( #1316 )
2026-03-04 23:47:38 +08:00
leejet
aaa8a51bd8
docs: update sd-cli/sd-server docs
2026-03-04 00:41:17 +08:00
leejet
ba35dd734e
refactor: introduce ggml_ext_zeros_like/ggml_ext_ones_like ( #1312 )
2026-03-04 00:36:52 +08:00
bssrdf
d41f5fff69
perf: improved flux attention qkv unpacking ( #1306 )
2026-03-04 00:36:32 +08:00
Korsar13
810ef0cf76
fix: reset weight adapter for models if no loras in request ( #1307 )
2026-03-04 00:34:07 +08:00
leejet
5792c66879
feat: support some non-standard Anima weight names ( #1305 )
2026-03-01 22:01:29 +08:00
Wagner Bruna
39d54702a6
feat: accept legacy image parameter on v1/images/edits ( #1270 )
2026-03-01 22:00:50 +08:00
Wagner Bruna
60889bc9a1
fix: correct sdapi LoRA file handling ( #1276 )
2026-03-01 21:57:06 +08:00
leejet
e64baa3611
refactor: reuse DiT's patchify/unpatchify functions ( #1304 )
2026-03-01 21:44:51 +08:00
leejet
cec4aedcfd
docs: add anima docs
2026-03-01 15:32:25 +08:00
rmatif
4cdfff5ff2
feat: add Anima support ( #1296 )
2026-03-01 15:23:18 +08:00
leejet
0752cc9d3a
fix: resolve image quality degradation issue ( #1297 )
2026-02-26 00:26:21 +08:00
Wagner Bruna and leejet
b314d80ad0
feat: turn flow_shift into a generation parameter ( #1289 )
...
* feat: turn flow_shift into a generation parameter
* format code
* simplify set_shift/set_parameters
* fix sd_sample_params_to_str
* remove unused variable
* update docs
---------
Co-authored-by: leejet <[email protected] >
2026-02-26 00:26:04 +08:00
leejet
c9cd49701a
fix: safely handle whitespace and consecutive newlines ( #1288 )
2026-02-19 20:54:42 +08:00
akleine
c5eb1e4137
fix: avoid black images if using an invalid VAE (for SDXL) ( #1273 )
2026-02-19 20:54:18 +08:00
leejet
636d3cb6ff
refactor: reorganize the vocab file structure ( #1271 )
2026-02-11 00:44:17 +08:00