Commit Graph

Select branches

Hide Pull Requests

arghh

avx512

batched

bench-memcpy

chess

ci/env

copilot/add-duplicate-text-removal

coreml-with-state

cuda-cublas-opts

diarization

distil-support

experiment/model-compression

fa-decoder

feature/debug-gradle-signing

fix-bench

fix-coreml-ane

fix-vzip

fix_vs_sdl2

gg/alloc-enc-results

gg/bench-fix-print

gg/benches-update

gg/chess

gg/ci-cuda-fix

gg/ci-fix-android

gg/ci-fix-windows

gg/ci-remove-label

gg/cuda-fix-mmvq

gg/cuda-no-async

gg/disable-cuda-graphs

gg/fix-external-encoder

gg/hipblas-fix

gg/make-fix-glob

gg/objc

gg/prompt-tokens

gg/reduce-ctx-use

gg/wchess

gg/whisper-short-audio-check

ggml-backend

ggml-backend-no-sched

ggml-conv

grammar-debug

guided

java-bindings

large-v3

llama-podcast

macros-cvt-fp16

master

metal

metal-and-alloc

nvblas

parallel-states

quantize-encoder

stream

sync-ggml-25-04-02-2

sync-ggml-25-05-07

sync-ggml-25-05-13

sync-ggml-25-09-30-2

sync-ggml-25-12-12

sync-ggml-25-12-17

sync-ggml-26-07-30

talk.llama-coreml

threads

timing

try-fix-abort

whisper-rc-v1.9.4

whisper-rc-v1.9.5

word-ts-2

#1001

#1002

#1003

#1010

#1012

#1015

#102

#1021

#1021

#1024

#1027

#1029

#1031

#1032

#1034

#1037

#1041

#1042

#1045

#1046

#1049

#1054

#1058

#1060

#1062

#1064

#1067

#107

#1074

#1074

#1077

#1081

#1086

#1086

#1092

#1097

#1097

#110

#1101

#111

#1110

#1111

#1112

#1113

#1114

#1115

#1118

#1118

#1120

#1124

#1128

#1129

#1130

#1131

#1134

#1136

#1137

#114

#1142

#1143

#1144

#1147

#1148

#115

#1154

#116

#1160

#1162

#1164

#1164

#1173

#1174

#1196

#1204

#1205

#1209

#121

#1210

#1211

#1212

#1214

#1216

#1217

#1218

#1220

#1224

#1227

#1228

#1229

#123

#1231

#1235

#1238

#124

#1243

#1247

#1250

#1251

#1253

#1254

#1255

#1261

#1261

#1263

#1264

#1265

#1267

#127

#127

#1270

#1275

#128

#1286

#1290

#1293

#1294

#1298

#130

#130

#1303

#1304

#1305

#1306

#131

#1310

#1313

#1317

#1330

#1334

#1335

#1345

#1349

#135

#1350

#1352

#1356

#1358

#136

#1362

#1364

#1368

#1370

#1375

#1375

#1380

#1381

#1381

#1382

#1389

#1400

#1404

#141

#1415

#1417

#1418

#1418

#1420

#1422

#1424

#143

#1432

#1434

#1440

#1441

#1442

#1444

#1445

#1452

#1455

#1455

#1456

#1457

#1458

#1459

#1462

#1466

#1467

#147

#1472

#1473

#1474

#1475

#1478

#1478

#1479

#1484

#1485

#1486

#1487

#1492

#1493

#1499

#1499

#150

#1500

#1500

#1501

#1505

#1519

#1521

#1522

#1523

#1524

#1524

#1529

#1530

#1533

#1534

#1535

#1539

#1541

#1544

#1545

#1546

#1547

#1548

#1549

#1549

#155

#1551

#1554

#1559

#1559

#1560

#1561

#1563

#1563

#1565

#1567

#1568

#1574

#1575

#1576

#1578

#1582

#1583

#1586

#1588

#1589

#1595

#160

#1602

#1604

#1604

#1605

#1606

#1607

#1615

#1617

#1627

#1627

#163

#1633

#1649

#1649

#1650

#1651

#1655

#1658

#1667

#1669

#1672

#1673

#1674

#1675

#1677

#1679

#1679

#1681

#1691

#1692

#1694

#1695

#170

#1701

#1703

#1704

#1713

#1714

#1716

#1717

#1725

#1727

#1728

#1729

#1735

#174

#1740

#1741

#1744

#1747

#1749

#175

#1750

#1753

#1754

#1755

#1758

#1763

#1764

#1765

#1768

#1768

#1772

#1774

#1778

#1781

#1785

#179

#1791

#1791

#1792

#1802

#1806

#1809

#1812

#1813

#1819

#1823

#1823

#183

#1833

#1833

#1838

#1839

#1840

#1841

#1841

#1842

#1850

#1854

#1854

#1857

#1859

#1860

#1861

#1863

#1865

#1871

#1872

#1874

#1878

#1888

#1889

#1890

#1891

#1895

#1897

#19

#1902

#1913

#1913

#1917

#1924

#1924

#1925

#1926

#1928

#1929

#193

#1932

#1933

#1938

#194

#1942

#1943

#1944

#1945

#1947

#195

#1952

#1952

#1953

#1964

#1965

#1966

#1969

#1969

#1970

#1973

#1973

#1978

#1980

#1981

#1982

#1983

#1990

#1990

#1994

#1997

#1998

#20

#2000

#2001

#2004

#2005

#2005

#201

#2012

#2019

#2020

#2024

#2025

#2026

#203

#203

#2043

#2044

#2045

#2048

#2049

#2054

#2058

#2063

#2068

#2068

#2069

#2070

#2071

#2071

#2072

#2073

#2075

#2075

#2080

#2086

#2088

#2090

#2094

#2095

#2095

#21

#2100

#2102

#2108

#2115

#2119

#2121

#2123

#2127

#2127

#2128

#2129

#2133

#2138

#2142

#2152

#2153

#2154

#2166

#2170

#2181

#2182

#2184

#2184

#2189

#2194

#2196

#2198

#2206

#2208

#2217

#222

#2220

#2227

#2231

#2232

#2234

#2235

#2236

#2237

#2238

#2239

#224

#2240

#2242

#2254

#2254

#2256

#2261

#2264

#2266

#2267

#2270

#2272

#2272

#2279

#228

#2288

#229

#2290

#2291

#2294

#2299

#23

#230

#2302

#231

#2311

#2324

#2330

#2336

#2339

#2342

#2343

#2346

#2350

#2358

#2360

#2367

#2369

#2376

#2382

#2383

#2384

#2386

#2387

#239

#2391

#2393

#2396

#24

#2401

#2406

#2406

#2407

#2410

#2414

#2416

#2417

#2419

#2424

#2425

#2427

#2429

#2431

#2432

#2432

#2433

#2440

#2443

#2444

#2449

#245

#2451

#2455

#2464

#2475

#2477

#2481

#2484

#2485

#2488

#2489

#2495

#2505

#2506

#2511

#2515

#2516

#2517

#2518

#2519

#252

#2523

#2525

#2528

#2529

#253

#2534

#254

#2543

#2546

#2547

#2548

#2549

#2550

#2551

#2555

#2560

#2561

#2562

#2567

#2569

#257

#2570

#2573

#2574

#2576

#2577

#2577

#2579

#2580

#2585

#2589

#2593

#2593

#260

#2604

#2608

#2611

#2613

#2617

#2623

#2624

#2625

#2629

#2633

#2634

#2634

#2635

#2637

#2638

#2639

#2641

#2642

#2643

#2648

#2649

#2653

#2654

#2656

#2659

#2663

#2664

#2670

#2674

#2676

#2683

#2684

#2686

#2687

#2690

#2690

#2691

#2691

#2692

#2693

#2694

#2694

#2699

#27

#2700

#2707

#2709

#271

#2711

#2716

#2718

#2728

#273

#2734

#2736

#2737

#274

#2745

#2749

#2756

#2759

#2760

#2769

#2769

#277

#2770

#2777

#2779

#2790

#2796

#2797

#2799

#28

#2800

#2800

#2816

#282

#2821

#2822

#2824

#2826

#2826

#2831

#2831

#2832

#2832

#2836

#2838

#2838

#284

#284

#2840

#2842

#2842

#2843

#2844

#2845

#2846

#285

#2851

#2853

#2855

#2858

#286

#2862

#2863

#2868

#287

#2873

#2875

#2876

#2877

#2878

#2879

#288

#2880

#2882

#2887

#2889

#2891

#2893

#2895

#2896

#29

#2900

#2902

#2904

#2905

#2908

#291

#2910

#2911

#2912

#2914

#2915

#2916

#2918

#2919

#2921

#2923

#2924

#2925

#2932

#2935

#2937

#2938

#2939

#294

#2941

#2942

#2943

#2945

#2946

#2947

#2948

#2949

#2951

#2952

#2953

#2955

#2956

#2958

#2959

#296

#2960

#2962

#2966

#2968

#2969

#2971

#2972

#2973

#2975

#2976

#2977

#2979

#298

#2981

#2985

#2986

#2987

#2988

#299

#2990

#2991

#2992

#2993

#2994

#2997

#2999

#3

#3000

#3001

#3002

#3004

#3005

#3006

#3007

#301

#3016

#302

#3021

#3022

#3024

#3025

#3027

#3028

#3029

#3031

#3033

#3038

#3042

#3043

#3044

#3045

#3050

#3052

#3054

#3054

#3055

#3056

#3057

#306

#3060

#3062

#3064

#3065

#3068

#3069

#3070

#3071

#3073

#3075

#3076

#308

#3082

#3083

#3084

#3085

#3086

#3087

#3090

#3097

#3098

#31

#3100

#3101

#3102

#3103

#3104

#3106

#3108

#3109

#3112

#3114

#3120

#3124

#3125

#3126

#3127

#3130

#3131

#3132

#3133

#3134

#3136

#3138

#3140

#3141

#3142

#3143

#3145

#3147

#3148

#3149

#3150

#3151

#3152

#3156

#3157

#3158

#3160

#3160

#3163

#3164

#317

#3170

#3171

#3172

#3173

#3175

#3177

#3178

#3179

#318

#3180

#3181

#3183

#3184

#3185

#3186

#3187

#3189

#319

#3190

#3191

#3192

#3193

#3195

#3196

#3197

#3199

#320

#3200

#3201

#3202

#3203

#3206

#3208

#3209

#3214

#3215

#3217

#3218

#3219

#322

#3220

#3221

#3222

#3223

#3223

#3229

#323

#3230

#3231

#3233

#3234

#3237

#3239

#324

#3241

#3242

#3243

#3244

#3245

#3246

#3247

#3251

#3255

#3257

#3257

#3261

#3262

#3264

#3265

#3266

#3268

#3270

#3272

#3273

#3274

#3274

#3275

#3276

#3277

#3281

#3282

#3283

#3284

#3287

#3288

#3289

#3291

#3291

#3292

#3294

#3296

#3298

#3298

#3300

#3301

#3307

#331

#3310

#3313

#3313

#3318

#3319

#3321

#3321

#3322

#3323

#3324

#3325

#3325

#3327

#3328

#3329

#3332

#3333

#3336

#3342

#3346

#3349

#3350

#3354

#336

#3363

#3365

#3369

#3371

#3372

#3374

#3374

#3378

#3378

#3379

#3381

#3383

#3387

#3387

#3389

#3394

#3395

#34

#340

#3401

#3401

#3406

#3408

#3409

#3412

#3412

#3416

#3417

#3417

#3419

#3422

#3422

#3423

#3423

#3425

#3426

#3428

#343

#343

#3430

#3430

#3433

#3433

#3436

#3437

#3438

#3439

#3441

#3442

#3443

#3445

#3447

#3448

#345

#3453

#3456

#3457

#3457

#346

#3461

#3462

#3462

#3463

#3466

#3467

#3468

#3468

#3469

#3470

#3471

#3471

#3472

#3473

#3474

#3474

#3477

#3478

#3482

#3483

#3484

#3485

#3485

#3487

#3488

#3489

#3489

#349

#3490

#3492

#3494

#3494

#3495

#3495

#3496

#3498

#350

#3502

#3502

#3503

#3505

#3505

#3506

#3506

#3507

#351

#3513

#3514

#3516

#3518

#3519

#3522

#3524

#3526

#3527

#3527

#3528

#3529

#3529

#353

#3532

#3533

#3534

#3535

#3540

#3540

#3543

#3543

#3555

#3555

#3557

#3558

#3559

#3563

#3564

#3565

#3565

#3566

#3568

#3569

#357

#3572

#3573

#3575

#3578

#3578

#3579

#3580

#3581

#3582

#3582

#3583

#3585

#3587

#3587

#3588

#3588

#3589

#3589

#359

#3590

#3591

#3592

#3593

#3594

#3594

#3598

#3598

#3599

#3599

#36

#3600

#3605

#3605

#3606

#3608

#3610

#3610

#3612

#3612

#3615

#3615

#3616

#3616

#3617

#3617

#3619

#3619

#362

#3621

#3624

#3625

#3625

#3626

#3626

#3630

#3632

#3632

#3633

#3636

#3637

#3637

#3638

#3638

#3639

#3641

#3641

#3644

#3646

#3647

#3649

#365

#3650

#3650

#3652

#3653

#3653

#3655

#3656

#3656

#3658

#366

#3660

#3660

#3661

#3661

#3665

#3668

#3670

#3671

#3672

#3675

#3677

#3678

#3679

#368

#3684

#3685

#3686

#3686

#3689

#369

#3691

#3691

#3693

#3696

#3696

#3699

#3699

#3703

#3704

#3704

#3705

#3707

#3709

#3709

#3710

#3711

#3713

#3714

#3714

#3715

#3716

#3716

#3719

#3719

#3727

#3730

#3732

#3732

#3735

#3742

#3742

#3748

#3748

#3751

#3755

#3755

#3756

#3757

#3757

#3762

#3763

#3764

#3765

#3766

#3767

#3768

#3768

#3769

#3770

#3770

#3775

#3776

#3780

#3780

#3781

#3782

#3783

#3784

#3785

#3786

#3789

#379

#3790

#3797

#3798

#38

#3803

#3805

#3808

#381

#3810

#3811

#3812

#3813

#3813

#3815

#3816

#3816

#3818

#3822

#3822

#3823

#3824

#3826

#3827

#3828

#3829

#383

#3830

#3831

#3832

#3833

#3834

#3835

#3837

#384

#3840

#3842

#3846

#3847

#3848

#3848

#3850

#3851

#3854

#3855

#3856

#3857

#3858

#3859

#3860

#3861

#3864

#3864

#3865

#3866

#3867

#3868

#3869

#3869

#387

#3870

#3870

#3872

#3874

#3874

#3875

#3878

#388

#3880

#3881

#3883

#3885

#3886

#3891

#3892

#3893

#3894

#3896

#3897

#3899

#390

#3900

#3904

#3904

#3905

#3906

#3907

#3908

#3909

#3909

#391

#3910

#3911

#3911

#3912

#3913

#3915

#3916

#3919

#3920

#3921

#3922

#3922

#3925

#3925

#3926

#3926

#3929

#3929

#3931

#3934

#3934

#3936

#3938

#3939

#3939

#3940

#3941

#3941

#3945

#3945

#3946

#3947

#3948

#3948

#3950

#3956

#3957

#3959

#3962

#3963

#3964

#3964

#3966

#3969

#3970

#3972

#3975

#3976

#3976

#3977

#3979

#398

#3980

#3980

#3981

#3982

#3985

#3987

#3987

#3989

#3990

#3991

#3994

#3996

#3997

#3998

#3999

#4000

#4001

#4002

#4003

#4004

#4005

#4006

#4007

#4008

#4009

#4010

#4012

#4013

#4015

#4016

#4017

#4019

#4020

#4020

#4021

#4021

#4022

#4023

#4023

#4025

#4028

#4028

#4029

#4031

#4032

#4033

#4033

#4034

#4035

#4038

#404

#4040

#4044

#4045

#4047

#4048

#4049

#4049

#4050

#4051

#4052

#4053

#4054

#4054

#4055

#4055

#4056

#4057

#4058

#4060

#4061

#4064

#4065

#4066

#4066

#4068

#4068

#4069

#4070

#4072

#4073

#4074

#4077

#4079

#4080

#4082

#4083

#4086

#4086

#4087

#4087

#4089

#4089

#409

#4091

#4095

#4095

#4096

#4096

#4097

#4097

#4099

#4099

#41

#4100

#4100

#4101

#4102

#4105

#4107

#4108

#4108

#415

#42

#424

#425

#43

#431

#435

#436

#439

#443

#444

#446

#451

#453

#454

#454

#455

#456

#459

#461

#462

#468

#473

#474

#476

#482

#484

#485

#486

#494

#495

#497

#500

#501

#502

#502

#503

#506

#515

#520

#523

#532

#534

#537

#538

#540

#542

#552

#563

#566

#569

#572

#576

#58

#583

#60

#600

#605

#613

#613

#615

#619

#624

#624

#626

#627

#628

#629

#629

#638

#640

#642

#645

#648

#649

#650

#650

#659

#659

#664

#668

#67

#677

#682

#685

#686

#687

#688

#697

#70

#704

#706

#710

#711

#712

#716

#718

#72

#720

#721

#725

#728

#733

#737

#739

#740

#755

#759

#760

#763

#764

#768

#77

#776

#78

#798

#81

#810

#811

#812

#815

#816

#832

#833

#834

#835

#836

#837

#842

#845

#853

#854

#862

#863

#867

#87

#871

#871

#874

#875

#883

#885

#890

#891

#893

#899

#902

#908

#910

#915

#926

#927

#931

#935

#939

#939

#94

#944

#95

#956

#964

#968

#968

#971

#971

#972

#995

0.0.5-3

0.0.6-1

1.0.3

1.0.4

1.1.0

1.4.1-1

1.4.1-2

1.5.2

b2250

b2251

b2252

b2253

b2254

b2257

b2258

b2259

b2260

b2261

b2262

b2263

b2266

b2267

b2268

b2269

b2270

b2271

b2273

b2274

b2275

b2276

b2279

b2280

b2339

b2340

b2341

b2342

b2348

b2349

b2350

b2351

b2352

b2353

b2362

b2364

b2365

b4938

b5127

b5130

b5454

danbev-java-jar-artifact-test

danbev-testing-xcframework-release

v1.0.4

v1.1.0

v1.1.1

v1.2.0

v1.2.1

v1.3.0

v1.4.0

v1.4.1

v1.4.2

v1.4.3

v1.5.0

v1.5.1

v1.5.2

v1.5.3

v1.5.4

v1.5.5

v1.6.0

v1.6.1

v1.6.2

v1.7.0

v1.7.1

v1.7.2

v1.7.2-pre

v1.7.3

v1.7.3-pre

v1.7.4

v1.7.4-pre-0

v1.7.4-pre-1

v1.7.5

v1.7.6

v1.8.0

v1.8.1

v1.8.2

v1.8.3

v1.8.4

v1.8.5

v1.8.6

v1.8.7

v1.9.0

v1.9.1

v1.9.2

v1.9.3

v1.9.4

v1.9.5

Mono Color

• b7b849500e Enables Windows ARM64 build with MSVC cl.exe (llama/28362) Sarah Wu 2026-09-28 01:07:27 -07:00
• 85f6926153 vulkan: fix wrong results when a mul_mat reads a slice of a larger cache (llama/28956) Pascal 2026-09-28 08:36:38 +02:00
• 51db47553c opencl: refine bin kernel loading condition (llama/29503) lhez 2026-09-27 03:08:39 -07:00
• 55a695da12 sycl: FWHT kernels for block widths above 512 (llama/29243) bri-prism 2026-09-27 03:08:19 -07:00
• 846af52607 CUDA: tune fp16 tile FlashAttention configs for head sizes 40-112 (llama/26289) Animesh 2026-09-27 15:37:37 +05:30
• 743f1ad286 HIP: Enable fattn-mma kernel on cdna for dkq > 256 for large batch sizes (llama/28907) uvos 2026-09-27 12:06:56 +02:00
• 14d1aa70e6 vulkan: fix argsort kernel selection for Adreno (llama/29469) Ruben Ortlam 2026-09-27 12:06:06 +02:00
• 24cf265ba6 RPC: use RDMA completion channel to not spin (llama/29440) Aman Gupta 2026-09-27 17:28:32 +08:00
• 7eea51881e hexagon: support tiled Q4_0 and Q8_0 GET_ROWS (llama/29511) kurquharandMax Krasnyansky 2026-09-26 23:02:37 -07:00
• fe06027d27 hexagon: support for backend sampler (llama/29502) Max Krasnyansky 2026-09-26 20:29:36 -07:00
• fc1ebfabf2 cuda: support Nemotron 3 Puzzle state size 96 for ssm scan (llama/28717) Anav Prasad 2026-09-26 13:05:52 -07:00
• dfe8fbd7e7 cuda: add F16 input to the FWHT (llama/29096) bri-prism 2026-09-26 12:36:09 -07:00
• 447a77510d ggml-cpu: tiled mul_mat for k-quants (llama/27851) jboothandGeorgi Gerganov 2026-09-26 13:38:57 +08:00
• bf1d787a7c opencl: add A8 Q8_0 non-MoE dp4a binary kernel (llama/29439) shaofeiqi 2026-09-25 20:41:47 -07:00
• 9731f6fdbf hexagon: find software divide calls using binary inspection tool (llama/29449) Trivikram Reddy 2026-09-25 21:43:38 -05:00
• 2d91497ae9 opencl: add bin kernel kernel_gemm_noshuffle_q5_k_f32_32b_trans_ila_a8_bin, kernel_gemm_noshuffle_q5_k_q8_1_dp4a_ila_a8_bin (llama/29401) shaofeiqiandLi He 2026-09-25 07:35:18 -07:00
• 669188ef02 Fixing the vulkan build issue of legacy GLSLC version that has no cooperativeMatrix API support (https://github.com/ggml-org/llama.cpp/issues/29373) (llama/29409) sliu39 2026-09-25 20:33:49 +08:00
• 9a919424e4 metal: FWHT kernels for block widths above 512 (llama/29095) bri-prismandGeorgi Gerganov 2026-09-25 02:15:33 -07:00
• 331d3c8aac sync : ggml Georgi Gerganov 2026-10-06 10:37:30 +03:00
• 4b0998ada3 metal : split fa kernels into per-dtype libraries (llama/29329) Georgi Gerganov 2026-10-06 10:36:44 +03:00
• bad9b54625 sync : ggml Georgi Gerganov 2026-10-06 10:34:37 +03:00
• afc10c26a4 llama : add llama_prec_policy + model-driven W4A4 path (llama/24364) 3 people 2026-09-25 08:36:35 +00:00
• d7e83b18cb HIP: bump HIP_VERSION requried for fp8 to avoid missing __hip_fp8_e4m3 support in 6.2 (llama/29231) uvos 2026-09-25 09:36:36 +02:00
• 26bfdc66ba rpc: include nb in the get_alloc_size cache key and floor the result at ggml_nbytes (llama/29283) Jess SullivanandGeorgi Gerganov 2026-09-25 03:35:09 -04:00
• 93c4308ae9 support sparse FA (llama/28796) Neo Zhang 2026-09-25 15:17:51 +08:00
• f7d5bf38fe musa: fix PH1 (MTT S5000) operator failures and build issues (llama/29193) R0CKSTAR 2026-09-25 15:16:29 +08:00
• 40c872b2fe CUDA: fuse RMS_NORM + SCALE into one kernel (llama/29393) InflexCZE 2026-09-25 07:22:43 +02:00
• 0458c9eb07 hexagon: add q5_k quant type support (llama/29123) Jhen-Jie Hong 2026-09-25 10:03:59 +08:00
• eea9aa6352 hexagon: use DMA for contiguous dim1 CONCAT (llama/29404) kurquhar 2026-09-24 19:03:41 -07:00
• 4550b9db9f metal : fix graph capture and handle empty graphs (llama/29390) Georgi Gerganov 2026-09-24 22:44:53 +03:00
• e1b93a513c metal : optimize sparse FA + clean-up (llama/29377) Georgi Gerganov 2026-09-24 22:44:36 +03:00
• 61032a2b44 hexagon: handle multi-sequence in concat_2d (llama/29344) Jhen-Jie Hong 2026-09-25 03:16:55 +08:00
• 845d1c02a8 hexagon: dynamic quantizer improvements (llama/29395) 4 people 2026-09-24 11:39:28 -07:00
• e5a88776af hexagon: support I32 CPY and CONT (llama/29379) kurquhar 2026-09-24 10:48:57 -07:00
• 5a9a3c069f cuda : add F16 kernel support for CONV_2D_DW (llama/29064) yomi 2026-09-25 00:38:46 +08:00
• cf0852d268 vulkan: int8 coopmat1 matmul implementation for AMD RDNA3 and RDNA4 (llama/27952) 3 people 2026-09-24 15:18:47 +02:00
• 6d0197cc48 ggml : bump version to 0.25.3 (ggml/1645) Georgi Gerganov 2026-09-24 21:01:29 +03:00
• a2b9a22e72 ggml : fix ubsan error in ggml_graph_nbytes (ggml/1644) Georgi Gerganov 2026-09-24 20:32:55 +03:00
• 7ab1529bec ggml : bump version to 0.25.2 (ggml/1642) Georgi Gerganov 2026-09-24 16:00:28 +03:00
• f938f31926 vulkan: handle misalignment in conv_2d and conv_3d (llama/29365) Ruben Ortlam 2026-09-24 14:08:26 +02:00
• bf8128ec03 vulkan: tune KHR cooperative matrix support for Adreno GPUs (llama/29328) Raman Shinde 2026-09-24 16:38:40 +05:30
• 3ef6a8320f cuda : add conv3d with implicit GEMM (llama/29137) leejet 2026-09-24 15:24:57 +08:00
• 6f2a84afdf hexagon: reject MUL_MAT_ID when src1 precision is F32 (llama/29348) Jhen-Jie Hong 2026-09-24 13:02:17 +08:00
• 0adffd58f6 opencl: add A8 Q6_K non-MoE dp4a binary kernel (llama/29057) shaofeiqi 2026-09-23 10:37:04 -07:00
• 60c0be6ac8 whisper : add check for ttype in whisper and parakeet (#4091) Vincenzo Reina 2026-10-02 14:21:13 +01:00
• 6e4ab854f6 cli : fix incorrect error code when files are failing (#4080) Vincenzo Reina 2026-09-28 19:35:45 +01:00
• d09f61a708 ci : cover GGML_BACKEND_DL in ubuntu-22-clang-arm64 (#4047) Abir Deol 2026-09-24 01:29:15 -07:00
• a664346ea5 sync : ggml Georgi Gerganov 2026-09-23 20:45:57 +03:00
• 84f080e977 ggml : bump version to 0.25.1 (ggml/1637) Georgi Gerganov 2026-09-23 20:10:09 +03:00
• 8f5ac2a62a CUDA: add a reserve to avoid spurious warning on older GCC builds (llama/29317) Aman Gupta 2026-09-24 00:52:40 +08:00
• 8917ea0432 metal: add the missing f32 x bf16 mul_mv variants (llama/28741) Pascal 2026-09-23 17:29:00 +02:00
• f48aebe2f1 CUDA: enable sparse-fa for dsv4 prefill (again) (llama/29298) Aman GuptaandPascal 2026-09-23 23:20:40 +08:00
• 6bc51cc745 metal : key the fa-vec tuned table by family instead of SKU (llama/29075) YiChen Lv 2026-09-23 19:23:15 +08:00
• 185213231b vulkan: add IQ4_XS MMQ/MMV matmul kernels (llama/28415) Piotr Wilkin (ilintar) 2026-09-23 09:00:06 +02:00
• ed1339a577 sync : ggml Georgi Gerganov 2026-09-23 11:41:19 +03:00
• 3e7723feef ggml : bump version to 0.25.0 (ggml/1635) Georgi Gerganov 2026-09-23 11:25:40 +03:00
• 711ef84442 common : fix for two functions when top_k exceeds the vocabulary size. (ggml/1633) Arianna Method 2026-09-23 10:45:06 +03:00
• f6b039ff0f sycl : fix compile warnings Georgi Gerganov 2026-09-23 10:44:07 +03:00
• 431ecf57cf ggml-meta: resolve multi buffer views (llama/29266) Ruben Ortlam 2026-09-23 07:35:24 +02:00
• e0ca36d049 cuda: top-k MoE should always fire (llama/28432) Aman Gupta 2026-09-23 13:26:05 +08:00
• d6075ffceb sycl : support new UT case for mul_mat_hadamard fp16 (llama/29218) Neo Zhang 2026-09-23 13:22:44 +08:00
• 6570b79e96 sycl: extend MMVQ GLU fusion, add rms_norm+scale and ssm_conv+silu fusions (llama/28931) Anant Shrivastava 2026-09-23 10:48:17 +05:30
• a1c7f973a0 sycl : support op get_rows_back, only support fp32/fp16 (llama/25266) Neo Zhang 2026-09-23 13:15:27 +08:00
• fbcb94d07f vulkan: hide internal symbols to prevent duplicate-dlopen state destruction (llama/29139) Erik Winter 2026-09-23 07:13:01 +02:00
• b18bad0d49 hex-dma: introduce direct-mapped DMA cache that is better suited for HVX FA mask handling (llama/29282) Max Krasnyansky 2026-09-22 15:20:08 -07:00
• e59366ba62 HIP : optimize IQ2/IQ3 (__vsub4 __vcmpne4) using SWAR (llama/27962) Felix Ye 2026-09-23 04:31:12 +08:00
• 736efcf9b0 opencl: add bin kernel kernel_gemm_noshuffle_q4_k_q8_1_dp4a_ila_a8_bin (llama/29056) shaofeiqi 2026-09-22 12:39:11 -07:00
• e900c88838 vulkan: add Intel Xe flash attention optimization kernels (2/3, Xe-LPG Plus/Xe2/Xe3) (llama/24406) Jiang, FishandLiu, Russell 2026-09-23 00:05:37 +08:00
• ad0058be6e metal : gate mul_mm_id src1 rescale behind ggml_prec (llama/29029) Michael de Gansandyomaytk 2026-09-22 17:32:28 +02:00
• ff565cf322 ggml : IQ1_M build prefix sums once per block (llama/28706) Bartowski 2026-09-22 15:54:45 +02:00
• 0409ed2628 Performance tune for gemma4-26b-a4b flash attention shape. (llama/28450) David M. Rogers 2026-09-22 09:43:29 -04:00
• dd67678efa opencl: add A8 Q4_0 non-MoE dp4a binary kernel (llama/29055) shaofeiqi 2026-09-21 23:14:00 -07:00
• 898392b724 hexagon: new HMX-optimized GATED_DELTA_NET (llama/29199) Max Krasnyansky 2026-09-21 14:49:52 -07:00
• 63412d34b8 metal : fix mask bounds in flash attention block pre-pass (llama/29220) Foad Abo Dahood 2026-09-21 20:31:56 +03:00
• 0e640a1801 cuda: fix sm_70 tile compilation error (llama/29224) lingyezhixing 2026-09-22 00:11:29 +08:00
• 9a7d43df1e ggml-cuda : convert contiguous tensors four elements at a time (llama/29155) Piotr Wilkin (ilintar) 2026-09-21 18:00:51 +02:00
• 86ef1b157f cuda : accelerate conv2d with implicit GEMM (llama/29135) leejet 2026-09-21 23:11:43 +08:00
• d18a422280 ggml : fix dimension and stride truncation in ggml_permute (llama/29227) leejet 2026-09-21 22:25:44 +08:00
• eb279cf65d sycl : support gated DSV4_HC_PRE and optional HC_POST comb matrix (llama/29132) cwriterandcwriter 2026-09-21 12:59:38 +02:00
• 0ecf57d7d7 sycl : pinned memory use right device context instead of 0 (llama/28895) Łukasz Ślusarczyk 2026-09-21 12:58:59 +02:00
• 34a4c124f7 CUDA: Follow up of #25635, refactoring FA shared smem swizzle (llama/28536) ynankani 2026-09-21 10:58:28 +00:00
• b7b1fe47e1 ggml-metal : simplify fusion pattern op list declaration (llama/29206) Georgi Gerganov 2026-09-21 12:37:24 +03:00
• 42c873a228 sycl : coalesce MKL-FA softmax loads instead of one work-item per row (llama/28918) Anant Shrivastava 2026-09-21 13:37:04 +05:30
• b0c514951e ggml-cpu: ARM Repack kernels for Q1_0 (llama/23492) pl752andCopilot Autofix powered by AI 2026-09-21 13:04:51 +05:00
• b50d5c319f hexagon: overhaul of buffer and DMA handling to support 64bit mappings + improvements (llama/29197) Max Krasnyansky 2026-09-21 01:00:28 -07:00
• 76d02a12b8 cuda : tune MMVQ to MMQ crossover for SM70 (Volta) (llama/28912) Yangyu ChenandJohannes Gäßler 2026-09-21 15:45:31 +08:00
• 87b6876bb8 metal : fix deprecation warnings from macOS 27 SDK (llama/29136) Niklas Wenzel 2026-09-21 09:44:40 +02:00
• 29e710c4a1 webgpu : add fused gdn + cpy (llama/28976) Masashi Yoshimura 2026-09-21 16:39:30 +09:00
• 4ef9fd70b4 CUDA: tune FA for Gemma 4 on Ampere or newer (llama/29152) Johannes Gäßler 2026-09-20 22:20:12 +02:00
• 984e400cc0 metal : support arbitrary hc in dsv4_hc_pre (llama/29169) Georgi Gerganov 2026-09-20 17:52:20 +03:00
• 3d949a3610 CUDA: enable sparse fa for qwen4 (llama/28770) Aman Gupta 2026-09-20 16:08:11 +08:00
• c77f6bbe35 metal: add F16 input to the FWHT (llama/29094) bri-prism 2026-09-19 21:57:30 -07:00
• 7f5ac73e23 hexagon: enable I32 GET_ROWS (llama/29116) Aparna M P 2026-09-19 22:18:31 +05:30
• 232d7189c3 hexagon: add support for GEGLU_QUICK (llama/29114) Aparna M P 2026-09-19 22:18:07 +05:30
• 099090cfd9 hexagon: enable support for TOP_K op (llama/29113) Aparna M P 2026-09-19 21:46:03 +05:30
• a00fa32e77 metal : add MoE and SSM_CONV fusion optimizations (llama/28948) Georgi Gerganov 2026-09-19 13:14:44 +03:00
• 47456f6f39 metal : fix FA support checks (llama/29122) Georgi Gerganov 2026-09-19 11:33:03 +03:00
• 4f1cce1427 metal : support qwen4exp hc ops (llama/29000) Georgi Gerganov 2026-09-19 11:27:30 +03:00
• 58844b02e1 cuda : fix CUB argsort corruption caused by in-place keys (llama/28389) 3 people 2026-09-19 00:32:52 -05:00
• 97dc017192 opencl: add support for bin kernel flash_attn_f32_f16_bin (llama/29046) dsproule 2026-09-18 16:32:31 -07:00