293 Commits

Author SHA1 Message Date
leejet
b56c68617d fix: avoid narrowing conversion in SigVQ patch embedding and format code 2026-09-21 01:27:31 +08:00
leejet
187b2561ea
feat: add native CUDA SageAttention support (#2005) 2026-09-21 01:16:59 +08:00
fszontagh
15f335daa5
feat: add LLaDA-Image support (#1968)
Co-authored-by: leejet <leejet714@gmail.com>
2026-09-21 01:14:20 +08:00
Wagner Bruna
b8248a869c
fix: avoid passing ggml logs as format strings (#2002) 2026-09-21 00:50:52 +08:00
Wagner Bruna
008ca5b492
feat: restore legacy fp8 handling when building with upstream ggml (#2001) 2026-09-21 00:48:30 +08:00
leejet
137f7409bb
feat: add Qwen Image 2.1 support (#1994) 2026-09-20 22:51:21 +08:00
leejet
1330cebae8
feat: support building with upstream ggml (#1999) 2026-09-19 22:25:01 +08:00
leejet
17860c0e45
perf: parallelize host tensor elementwise and broadcast ops (#1998) 2026-09-19 21:46:16 +08:00
leejet
275ab58e01
perf: reduce CPU overhead in graph execution and sampling (#1997) 2026-09-19 18:32:58 +08:00
Fabrice Aneche
d32b4e893b
fix: prevent clip_preprocess center crop from exceeding the resized image (#1995) 2026-09-19 18:08:34 +08:00
leejet
3e037a81e4
perf: accelerate VAE direct 3D convolutions (#1996) 2026-09-19 17:50:01 +08:00
leejet
adcac69650
perf: pad small attention heads to 64 for MMA Flash Attention (#1992) 2026-09-18 23:59:58 +08:00
leejet
656a1354c3
refactor: remove obsolete unused tensor filtering (#1984) 2026-09-18 23:41:37 +08:00
Lin Xuhao
269e726015
fix: honor flash attention flag in LLM text encoder attention (#1987) 2026-09-18 23:41:24 +08:00
leejet
cc515a01f9
perf: eliminate temporary allocations in Philox rounds (#1982) 2026-09-17 02:09:12 +08:00
leejet
3161505fe8
fix: remove vision_model. from ununsed tensors (#1983) 2026-09-17 02:08:24 +08:00
leejet
59c23bce0d
fix: use tokenizer-specific pre-tokenization rules (#1975) 2026-09-15 02:37:52 +08:00
Wagner Bruna
07a85c74cb
feat: support Brownian tree noise in all noise injection samplers (#1899) 2026-09-15 02:37:31 +08:00
leejet
f9ddc0f388
refactor: require external Gemma 2 and GPT-OSS tokenizers (#1974) 2026-09-15 01:25:13 +08:00
leejet
4964abdfc5
feat: support external Hugging Face tokenizer JSON files (#1973) 2026-09-15 00:27:39 +08:00
Санька Четвёртый
42d6c0ab92
feat: Add generation parameters into video metadata (#1901)
Co-authored-by: leejet <leejet714@gmail.com>
2026-09-14 00:01:09 +08:00
leejet
5a5400bf0c
fix: resolve MSVC narrowing conversion warnings (#1969) 2026-09-13 23:45:54 +08:00
fszontagh
ca37fad89a
fix: validate vision projector output dim against LLM hidden size (#1918) 2026-09-13 23:42:46 +08:00
George
0bd72f075a
feat: add Wan2.2 S2V (audio+img-to-video) support (#1925)
Co-authored-by: leejet <leejet714@gmail.com>
2026-09-13 23:33:50 +08:00
fszontagh
4a7da26b73
fix: bound plain-text runs in parse_prompt_attention regex (#1919) 2026-09-13 23:29:19 +08:00
leejet
9a977388a8
fix: guard GPU memory capacity and propagate encoding failures (#1958) 2026-09-13 21:35:46 +08:00
leejet
44dd13716d
feat: preserve explicit backend assignments during auto-fit (#1967) 2026-09-13 17:40:38 +08:00
leejet
7f410a3793
feat: add linear and attention scale overrides (#1964) 2026-09-12 01:41:28 +08:00
leejet
5ebce93342
fix: reuse graph plans when scale parameters change (#1963) 2026-09-12 01:04:25 +08:00
Maphist0
7f986a9d73
feat: add SenseNova U1.5 support (#1935) 2026-09-12 00:39:59 +08:00
fszontagh
e06b205384
feat: expose the loaded model version name through the public API (#1962) 2026-09-11 23:39:56 +08:00
LED-M
e95ab96997
fix: preserve BF16 embedding weights for get_rows (#1959) 2026-09-11 23:09:45 +08:00
Hmission
b68d58624d
fix: enable VAE decode tiling fallback without auto-fit (#1932) 2026-09-11 01:34:56 +08:00
leejet
14eddb32b1
refactor: split generation pipeline out of stable-diffusion.cpp (#1957) 2026-09-11 00:57:14 +08:00
stduhpf
469fc49bb7
docs: reflect GGML_MAX_NAME value change in rpc docs (and in ggml_extend assert) (#1950) 2026-09-10 23:55:02 +08:00
leejet
6b47fec013
refactor: unify model source and weight lifecycle management (#1956) 2026-09-10 23:53:48 +08:00
Wagner Bruna
d04e8950c1
fix: correct SDXL embeddings loading (#1939) 2026-09-08 00:23:08 +08:00
nan
9cdb6b6fc0
fix: preserve K-quantized embedding weights (#1936) 2026-09-07 23:50:16 +08:00
leejet
31ab2b2e08
refactor: split ggml extensions and move implementations to cpp files (#1945) 2026-09-07 23:34:45 +08:00
leejet
d8fb10c029
fix: reuse graph cut plans across CFG passes (#1943) 2026-09-07 00:39:49 +08:00
leejet
80bac2d5fc
feat: enable single-GPU auto-fit with tiered parameter placement (#1942) 2026-09-07 00:18:42 +08:00
leejet
dbb611264e
feat: add verbose logging and log-level selection (#1941) 2026-09-06 23:14:03 +08:00
leejet
462d675018
refactor: unify runner lifecycles and weight residency (#1940) 2026-09-06 22:30:45 +08:00
assouan
6c57cc3b38
feat: prefetch streamed layers during compute (#1905)
Co-authored-by: leejet <leejet714@gmail.com>
2026-09-06 16:35:30 +08:00
leejet
6b3edaaf32
feat: generalize temporal tiling across video VAEs (#1926) 2026-08-31 00:11:38 +08:00
leejet
40e605f3f1
fix: use carrier sampling for MiniMax H3 audio (#1924) 2026-08-30 21:57:02 +08:00
vmobilis
d9b6e27e9f
feat: additional --preview-interval values (#1915)
Co-authored-by: leejet <leejet714@gmail.com>
2026-08-30 21:47:11 +08:00
leejet
2540a4fc25
feat: use backend-native FP8 matmul when supported (#1916) 2026-08-30 20:19:56 +08:00
jk212h20
dc4000d9f8
fix: correct MiniMax H3 audio Euler steps (#1908) 2026-08-30 20:19:28 +08:00
jk212h20
c797899732
fix: correct MiniMax H3 reference audio encoding (#1886) 2026-08-30 19:33:59 +08:00