nnfusion

Граф коммитов

Выбрать ветку

Скрыть запросы на слияние

LadderLLM

antares_dev

cocktailer_artifact

control_flow

diffusion

donglinb/dsl

dsl_fusion

dsl_kernel

fix-cub

fractaltensor_artifact

gnn

half_type

jlxue-patch-1

jlxue-patch-2

jlxue/fix_url

jxue/diffuser

jxue/op_fix

jxue/opt

jxue/python_interface

jxue/swin

jxue/symbolic

l1

lingm/antares_const_folding

lingm/op_fix

lingm/python_train_inference

lingm/rocm_kernel_db

lingm/wrf

lingm/xbox

main

master

nnf_diffuser

onnx_bert

osdi2023welder

osdi20_artifact

osdi22_artifact

revert-198-fix_tensor_name

rocm

roller

roller_ae_fix

roller_ae_working

roller_e2e_fix

symbolic

v0.3_rc

wenxh/benchmark

wenxh/diffusers

wenxh/div_add_softmax_fusion

wenxh/fix_sym

wenxh/onnx_unittest

wenxh/op_fix

wenxh/opt

wenxh/srop

wenxh/sym

wenxh/symbolic

wenxh/xbox_dataclass

wenxh/xbox_onnx_test

workflow

xbox

xbox_fix

xbox_test

yuqxia/add_purify_graph_pass

yuqxia/diffuser

yuqxia/fix_bug

yuqxia/fix_conv_fusion

yuqxia/fix_register_fusion_pass

yuqxia/merge

yuqxia/osdi20_artifac

yuqxia/prepare

yuqxia/validateop

yuxgu/issue-fix

yuxgu/virconf

zimiao/fix_memory_leak

#1

#10

#100

#101

#102

#103

#104

#106

#107

#109

#111

#112

#113

#12

#126

#13

#130

#133

#134

#137

#138

#139

#14

#140

#141

#142

#143

#145

#146

#148

#149

#15

#150

#152

#154

#155

#158

#16

#161

#165

#167

#168

#169

#17

#170

#171

#173

#174

#178

#18

#180

#181

#182

#183

#184

#186

#187

#188

#189

#193

#196

#198

#199

#2

#20

#202

#203

#205

#206

#209

#21

#210

#213

#214

#215

#216

#219

#22

#220

#221

#222

#223

#224

#225

#227

#229

#230

#231

#232

#233

#234

#235

#236

#237

#238

#239

#240

#240

#241

#242

#243

#244

#245

#249

#25

#253

#255

#256

#257

#258

#259

#259

#267

#269

#270

#271

#271

#272

#273

#274

#275

#276

#278

#280

#283

#286

#288

#289

#29

#291

#292

#293

#294

#297

#298

#299

#3

#300

#302

#303

#304

#306

#307

#308

#309

#310

#313

#315

#316

#317

#32

#320

#321

#322

#323

#329

#33

#330

#331

#333

#334

#335

#336

#338

#339

#34

#340

#341

#342

#343

#345

#346

#348

#349

#35

#351

#353

#353

#361

#363

#364

#364

#365

#366

#37

#372

#374

#375

#377

#378

#379

#380

#381

#382

#383

#384

#389

#39

#390

#391

#392

#393

#394

#395

#396

#397

#398

#4

#400

#401

#402

#403

#406

#407

#408

#409

#410

#410

#411

#413

#414

#416

#418

#418

#419

#420

#421

#422

#423

#424

#426

#427

#428

#429

#43

#430

#432

#433

#435

#437

#438

#439

#440

#444

#448

#449

#451

#451

#453

#453

#454

#455

#456

#458

#459

#460

#461

#465

#465

#466

#467

#47

#470

#471

#472

#473

#474

#475

#476

#477

#478

#479

#480

#481

#482

#483

#484

#485

#486

#487

#488

#489

#490

#491

#492

#493

#494

#495

#496

#497

#498

#499

#5

#500

#501

#502

#503

#504

#506

#509

#510

#511

#512

#513

#514

#514

#515

#516

#518

#522

#523

#524

#528

#531

#54

#56

#58

#6

#60

#62

#63

#7

#71

#8

#80

#81

#83

#84

#85

#9

#90

#91

#92

#94

#95

#96

#97

v0.1

v0.2

v0.3

v0.4

f2d7df38f5 add more guides heheda 2023-04-19 10:43:05 +0800
8266a71f5b fix bug in scripts heheda 2023-04-18 20:59:32 +0800
77fbf5b499 change path heheda 2023-04-18 15:06:53 +0800
ac6bd1feb6 copy file from ControlFlow repo heheda 2023-04-18 14:31:53 +0800
8f09a102c2 update ir yuqxia 2023-04-17 11:33:27 +0000
df3126ce44 Add artifect files. Shi Yining 2023-04-17 08:42:25 +0000
810a2e6cd8 add mem eff attn basic yuqxia 2023-04-17 08:45:14 +0000
cda34737a8 Update profile scripts for baselines. Shi Yining 2023-04-17 07:50:29 +0000
845c52cf48 add __syncthreads() to cf kernels heheda 2023-04-17 15:07:46 +0800
b03e0a97a4 support layout of layoutdot LeiWang1999 2023-04-15 23:54:15 -0800
7b605e3726 add dot permutation pass LeiWang1999 2023-04-15 07:22:11 -0800
26685bf03c Add a docker file. Shi Yining 2023-04-14 11:32:47 +0000
f1383415bb wrap python part with ifdef heheda 2023-04-14 13:58:34 +0800
6b681e3a4d bug fix .. LeiWang1999 2023-04-11 23:52:19 -0800
48bc398e93 apply code style, revert typedef_int change yuqxia 2023-04-12 06:58:04 +0000
90df32d7d0 resolve conflict yuqxia 2023-04-12 06:46:28 +0000
ca473321f5 fix bug yuqxia 2023-04-12 06:44:46 +0000
b62ce8400a re-type the CUDA_ARCH String LeiWang1999 2023-04-11 22:36:29 -0800
7eb413f166

fix bug (#515) Yuqing 2023-04-12 12:19:55 +0800
e24e0c6833 fix bug yuqxia/fix_register_fusion_pass yuqing 2023-04-12 11:26:03 +0900
4cb6ce3f07 fix bugs of register fusion pass LeiWang1999 2023-04-11 04:23:57 -0800
556158b6c2 add support for int16_t load (bloom fp16 model) LeiWang1999 2023-04-11 04:23:35 -0800
f0c358d685 fix blockfusion sync problem heheda 2023-04-10 16:27:51 +0800
5ccabaca7d Update Volta Tensorcore template. Shi Yining 2023-04-10 04:10:00 +0000
1859e9f33e Rename python package name to welder. Shi Yining 2023-04-07 13:20:37 +0000
a73a8b9cfe Update reduce step policy. Shi Yining 2023-04-07 01:35:56 +0000
92bd514054 Update README.md Shi Yining 2023-04-07 01:35:45 +0000
9ec9afd68a Some bug fix. Shi Yining 2023-04-06 07:55:31 +0000
d55331f2a5 Update readme install command Shi Yining 2023-04-05 00:41:55 +0000
e930bf9f99 Profile kernel in a seperate process Shi Yining 2023-04-05 09:22:18 +0900
5d6413f8e0 Fix local fuse when op is skipped Shi Yining 2023-04-04 23:26:22 +0900
df5b963d7f fixup Shi Yining 2023-04-03 08:46:28 +0000
65c6f4f5d7 search unroll width heheda 2023-04-03 16:08:13 +0800
f5e67390e5 dump kerneldb requests heheda 2023-04-03 15:49:18 +0800
60a287afab fix depunit bug in loop heheda 2023-04-03 11:48:18 +0800
c8cce319e4 remove cudadevicereset heheda 2023-04-03 11:48:02 +0800
6c403e256d fixup Shi Yining 2023-04-02 04:48:34 +0000
d210a5fde4 Add cache tensor to policy Shi Yining 2023-03-31 05:00:59 +0000
547136da2e reorganize parameters heheda 2023-03-30 21:02:00 +0800
724cbf5693 add to nhwc pass yuqxia 2023-03-30 06:13:30 +0000
8ae6e90877

disable external result memory (#513) Ziming Miao 2023-03-29 12:57:29 +0800
d92928d327 disable external result memory zimiao 2023-03-29 12:55:15 +0800
ecf47cca8f add IfSingle operator heheda 2023-03-28 23:15:51 +0800
f23589621f merge rocm code heheda 2023-03-27 16:58:33 +0800
a25b432f22 [Experimental] Implement local fusion case Shi Yining 2023-03-27 06:26:53 +0000
581f0e5eb0 Add inverse shape inference. Shi Yining 2023-03-26 06:46:54 +0000
676a01a83a Fix an engine cache bug Shi Yining 2023-03-23 10:43:35 +0000
af213acc17 fix check yuqxia 2023-03-22 04:35:08 +0000
f86906f59c Merge branch 'lingm/xbox' of https://github.com/microsoft/nnfusion into lingm/xbox Lingxiao Ma 2023-03-17 15:49:44 +0800
da721d78c6 merge origin/xbox branch Lingxiao Ma 2023-03-17 15:43:33 +0800
37316c75d7 Fix mma policy infer padded shared memory size Shi Yining 2023-03-17 02:19:28 +0000
6eb1445f56 Add test operators and MMA tests Shi Yining 2023-03-17 10:58:33 +0900
31f604c474 Add condition remove pass after vectorizing Shi Yining 2023-03-17 10:09:56 +0900
cd78f1adb4 merge welder yuqxia 2023-03-16 09:27:30 +0000
2aa530bac4 Update TIR scheduler Shi Yining 2023-03-15 10:48:30 +0000
40bea8caa8

meet int16 request in fp16 bloom model (#512) Lei Wang 2023-03-14 16:57:55 +0800
f338e372b6 Update te shape inference Shi Yining 2023-03-14 03:12:25 +0000
b382152d25 Add Nimble. Shi Yining 2023-03-11 14:26:41 +0000
6cfa2be0b9 meet int16 request in fp16 bloom model LeiWang1999 2023-03-11 11:42:48 +0000
fc58e7756a add transpose-matmul fusion in subgraph_fusion Lingxiao Ma 2023-03-10 16:26:38 +0800
9d3b56cb30 add transpose-matmul fusion in subgraph_fusion Lingxiao Ma 2023-03-10 16:08:35 +0800
9253962b2c

Fix FileNotFoundError when tuning_steps is 0 (#511) Hongzhou Liu 2023-03-10 13:04:52 +0800
e55e3b0a7d Fix FileNotFoundError when tuning_steps is 0 Hongzhou Liu 2023-03-08 00:03:03 +0800
77fd82bcdb Move handle for proxy output to schedule Shi Yining 2023-03-06 06:22:07 +0000
bdd695fead add arch, policyv2 roller yuqing 2023-03-07 16:30:16 +0900
4d2abb829b merge xbox yuqing 2023-03-07 16:26:26 +0900
e703e344ce added HLSL test pipeline on windows donglinb 2023-03-06 22:41:07 +0800
23fa719ea5 release loaded DLL in destructor donglinb 2023-03-06 15:56:39 +0800
1682bd845c Add check for vector load Your Name 2023-03-04 10:42:39 +0900
8b33cc3a52

Merge branch 'microsoft:xbox' into xbox donglinb 2023-03-01 18:23:13 -0800
dc8389bc5e Fix tir mma schedule Your Name 2023-03-02 00:47:33 +0900
9256a30f72 Update header Shi Yining 2023-03-01 12:34:41 +0000
a06d5e7b2b Update policy Shi Yining 2023-02-28 11:47:19 +0000
317b2b3b76 Update policy stride propogate code. Shi Yining 2023-02-28 04:43:04 +0000
e26cc77cb7

Flags to disable emitting cudaSetDevice call (#509) Ziming Miao 2023-02-27 19:48:13 +0800
3447cf49ef flags to disable emitting cudaSetDevice Ziming Miao 2023-02-27 20:46:52 +0900
1c69df6807 disable const folding in onnx exportor Ziming Miao 2023-02-27 20:32:05 +0900
0e494f6fc5

Yuqxia/diffuser (#506) Yuqing 2023-02-27 14:21:16 +0800
9b96aa80be convert int64_t to long long yuqxia/diffuser yuqing 2023-02-27 15:07:03 +0900
e45d274c84 add cutlass submodule Ziming Miao 2023-02-27 13:31:33 +0900
14cfed9629 delete cutlass index Ziming Miao 2023-02-27 13:29:04 +0900
5121aa1757 Update V100 policy Shi Yining 2023-02-24 15:45:56 +0000
b38e16f864 Refine extract subgraph logic Shi Yining 2023-02-24 14:54:39 +0000
4e5603207a Fix a tir op bug in te shape inference Shi Yining 2023-02-23 08:48:15 +0000
257f5337b8 Add log op fix Shi Yining 2023-02-23 08:37:46 +0000
e168101f6c Skip parsing ir for skipped nodes Shi Yining 2023-02-23 08:37:34 +0000
e40f11737e Switch shape inference from antares to te. Shi Yining 2023-02-22 08:44:56 +0000
211b18dd81 Add TIR pass for mma codegen Shi Yining 2023-02-22 08:33:06 +0000
272d07e86c fix gatherv2 Yuqing 2023-02-22 07:57:27 +0000
8546c1010e

Merge branch 'microsoft:xbox' into xbox donglinb 2023-02-20 02:49:36 -0800
cb2ffaa37b Remove unused files Shi Yining 2023-02-19 07:57:12 +0000
46758fa33b fixup Your Name 2023-02-19 16:18:54 +0900
54e5e7a8fc Add mma schedule for sm80 Shi Yining 2023-02-19 03:10:50 +0000
6dc7c0799b fixup Shi Yining 2023-02-18 05:02:26 +0000
1e88dcabd4

Invoke all ONNX tests (#502) wx 2023-02-17 12:41:03 +0800
a918264fc5 Implement TIR MMA scheduler Shi Yining 2023-02-16 05:25:18 +0000
a9007e5f91 update antares wrapper zimiao 2023-02-15 20:33:58 +0800
bbdbfb3f10 update Reduce ops for opset18 and ground_truth results xbox_test Jilong Xue 2023-02-14 20:38:46 +0800
dfc26dd082 Merge remote-tracking branch 'origin/xbox' into xbox_test Jilong Xue 2023-02-14 13:08:01 +0800
fd25765ab7 Refactor schedule and build code structure Shi Yining 2023-02-12 07:17:08 +0000