 leejet
|
b9d7176bf8
|
feat: unify text encoder selection across H3 variants (#2340)
|
2026-11-13 00:23:54 +08:00 |
|
|
Artem
|
8b1e9df27d
|
feat: allow the P-H3-Image text encoder for MiniMax-H3-Image (#2337)
|
2026-11-12 23:20:18 +08:00 |
|
|
Wagner Bruna
|
766ce0ec4e
|
perf: replace qkv materialized chunks with views (#2335)
|
2026-11-12 23:10:48 +08:00 |
|
|
leejet
|
32d516c019
|
feat: add MiniMax-H3-Image support (#2324)
|
2026-11-11 01:24:25 +08:00 |
|
|
losewayy
|
f8417d02e1
|
perf(cuda): fuse RMSNorm and RoPE in Sol-H3-VAE (#2323)
|
2026-11-11 01:06:53 +08:00 |
|
|
leejet
|
465adf389f
|
fix: avoid structured binding capture in MiniMax H3 SOL attention
|
2026-11-11 01:02:14 +08:00 |
|
|
leejet
|
0678d04ae7
|
fix(vae): order Sol attention and FFN writes to reduce VRAM
|
2026-11-10 23:56:53 +08:00 |
|
|
leejet
|
ddc367ab5d
|
refactor: unify runner execution and weight residency (#2321)
|
2026-11-10 23:45:01 +08:00 |
|
|
leejet
|
56070ea034
|
refactor: replace QKV split helpers with ggml_ext_chunk (#2322)
|
2026-11-10 23:38:47 +08:00 |
|
|
losewayy
|
0bb132831f
|
feat: support Sol-H3-VAE decoding for MiniMax-H3 (#2317)
|
2026-11-10 23:29:17 +08:00 |
|
|
losewayy
|
e1ca5e8bc7
|
perf(cuda): fuse bias into INT8 CUTLASS epilogues (#2320)
|
2026-11-10 23:08:05 +08:00 |
|
|
leejet
|
a3c49b13be
|
refactor: unify ggml operation helpers around runner context (#2314)
|
2026-11-10 01:58:04 +08:00 |
|
|
Windyskr
|
46da352af0
|
fix: size layer-split ranges in proportion to device capacity (#2309)
|
2026-11-09 23:40:26 +08:00 |
|
|
Windyskr
|
008b9a20ca
|
fix: apply MiniMax-H3 Diffusers LoRA token_refiner weights (#2306)
|
2026-11-09 22:21:46 +08:00 |
|
|
Windyskr
|
cfdf3cf037
|
fix: merge LoRAs in F32 when convert changes the tensor type (#2302)
|
2026-11-09 22:16:04 +08:00 |
|
|
Windyskr
|
d22dcda1cf
|
perf: cache the device type in sd_backend_is_cpu (#2307)
|
2026-11-09 22:13:35 +08:00 |
|
|
Wagner Bruna
|
f977addf06
|
fix: disable batch_cfg for Metal when building with upstream ggml (#2311)
|
2026-11-09 22:10:12 +08:00 |
|
|
leejetandMaxime Dréan
|
fc350be2ce
|
feat(server): report generation stages independently of previews (#2303)
Co-authored-by: Maxime Dréan <maximedrean@proton.me>
|
2026-11-09 02:32:27 +08:00 |
|
|
leejet
|
b43ce4badd
|
fix: preserve Krea2 text-fusion names in Kohya LoRAs (#2301)
|
2026-11-09 02:11:28 +08:00 |
|
|
leejet
|
cd5594e9f9
|
feat: add SeedVR2 image and video upscaling (#2300)
|
2026-11-09 01:13:57 +08:00 |
|
|
leejetandVideo Rebirth
|
80c4707c53
|
feat: support HyperFlow H3 Flash (#2299)
Co-authored-by: Video Rebirth <opensource@videorebirth.com>
|
2026-11-09 00:43:02 +08:00 |
|
|
leejet
|
3d315c47d4
|
feat: add Metal INT8 ConvRot support (#2297)
|
2026-11-08 22:28:35 +08:00 |
|
|
DrMkdaddy
|
b43af40461
|
fix: use patch token count for the Flux scheduler (#2293)
|
2026-11-08 21:37:06 +08:00 |
|
|
leejet
|
0b3b98d7b8
|
fix: avoid CPU fallback based on backend allocation chunk size (#2294)
|
2026-11-08 16:24:53 +08:00 |
|
|
leejet
|
de00bafc3d
|
feat: add configurable attention chunk budget for MiniMax-H3 (#2292)
|
2026-11-08 15:50:25 +08:00 |
|
|
Zihao
|
7bcd4ce8ce
|
fix: clean LoRA multiplier and prompt spacing in image metadata (#2283)
|
2026-11-08 15:23:11 +08:00 |
|
|
leejet
|
5f8c19f98d
|
perf: reduce fused compute buffers (#2290)
|
2026-11-08 12:28:44 +08:00 |
|
|
leejet
|
9234ad822a
|
feat: add RelayVSR video upscaling with CUDA window attention (#2285)
|
2026-11-08 00:09:48 +08:00 |
|
|
leejet
|
b1f5a51a1e
|
fix: chunk MiniMax-H3 QKV projections within backend buffer limits (#2286)
|
2026-11-08 00:04:37 +08:00 |
|
|
Daniele
|
fdfc913c51
|
feat: ControlNet Union support for SDXL (#2284)
|
2026-11-07 23:33:41 +08:00 |
|
|
leejet
|
5a1ba5d895
|
perf(vulkan): accelerate packed INT4 with cooperative matrices (#2280)
|
2026-11-07 00:58:26 +08:00 |
|
|
leejet
|
f5df5f8b84
|
perf: accelerate packed INT4 CUDA inference with CUTLASS (#2278)
|
2026-11-06 01:59:05 +08:00 |
|
|
leejetandnoctrex
|
7952693004
|
feat: support packed INT4 convrot models (#2276)
Co-authored-by: noctrex <noctrex@gmail.com>
|
2026-11-06 00:35:58 +08:00 |
|
|
leejetandmirek190
|
58bb3c1b1c
|
perf(vulkan): accelerate MOSS normalization and SiTU staging (#2273)
Co-authored-by: mirek190 <mirek190@gmail.com>
|
2026-11-05 03:01:31 +08:00 |
|
|
farawayso
|
d1358945dc
|
feat: MiniMax-H3 latent hires second pass (#2249)
|
2026-11-05 02:59:04 +08:00 |
|
|
leejet
|
3e92bf990a
|
feat: add P-H3-Image generation and editing support (#2272)
|
2026-11-04 23:24:51 +08:00 |
|
|
leejetandmirek190
|
984eb0686c
|
perf(cuda): fuse MOSS modulation and direct RoPE F16 output (#2269)
Co-authored-by: mirek190 <mirek190@gmail.com>
|
2026-11-04 02:01:41 +08:00 |
|
|
leejet
|
d30854463d
|
perf: fuse Z-Image GPU operations and optimize BF16 GEMM (#2265)
|
2026-11-03 23:59:21 +08:00 |
|
|
Piotr Wilkin (ilintar)
|
6225bf8c88
|
perf: view qkv slices for moss instead of slice+reshape (#2232)
|
2026-11-03 23:46:35 +08:00 |
|
|
Mirek
|
f8ea2e162c
|
perf: use native SiTU for MOSS with composite fallback (#2210)
|
2026-11-03 23:41:09 +08:00 |
|
|
leejet
|
b4d6b83af6
|
fix: support integer LoRA scalar conversion to FP32 (#2263)
|
2026-11-03 00:07:40 +08:00 |
|
|
cumal
|
250ea6e385
|
fix: layer split across cuda (#2262)
|
2026-11-03 00:03:43 +08:00 |
|
|
leejet
|
290148d2e3
|
feat: add custom modulate operator and ConvRot fusion (#2258)
|
2026-11-02 01:19:28 +08:00 |
|
|
leejet
|
133087deb8
|
fix: prefer discrete GPUs in auto-fit device selection (#2259)
|
2026-11-02 01:09:36 +08:00 |
|
|
leejet
|
62c54439c8
|
fix: include missing cassert headers in SageAttention CUDA code
|
2026-11-02 00:17:49 +08:00 |
|
|
leejet
|
fe2682ba16
|
fix: retry Vulkan capacity checks after memory reclamation (#2256)
|
2026-11-01 22:55:56 +08:00 |
|
|
leejet
|
5356982584
|
fix: chunk MiniMax-H3 MLP to respect backend buffer limits (#2257)
|
2026-11-01 22:08:29 +08:00 |
|
|
leejet
|
45a63ac460
|
refactor: move model-specific operators and fusions into backend extensions (#2253)
|
2026-11-01 21:18:37 +08:00 |
|
|
leejet
|
0a067120b3
|
perf: optimize Vulkan kernels, graph fusion, and workspace reuse (#2251)
|
2026-10-31 14:55:58 +08:00 |
|
|
leejet
|
e239e4b0fd
|
refactor: replace manual gated activations with native ggml GLU (#2250)
|
2026-10-30 01:32:46 +08:00 |
|