Commit Graph

  • 8a02847b62 feat: 支持导出模型带数据 YdrMaster 2023-09-20 13:35:40 +0800
  • 86e6dc7fd9 feat: 导出张量时使用 numpy 结构以兼容任何数据类型 YdrMaster 2023-09-20 13:14:58 +0800
  • 054e8b4990 feat: 支持 runCpu YdrMaster 2023-09-20 11:02:04 +0800
  • 166e0c0a08 feat: 支持从前端取得折叠的输出 YdrMaster 2023-09-20 09:49:17 +0800
  • 4c1e32cac8 merge master panzezhong 2023-09-20 09:21:24 +0800
  • 8086f3d34b fix: 必须保存 GlobalOutputs 以维护顺序访问 YdrMaster 2023-09-20 09:10:54 +0800
  • 6cd08abad9 feat: 直接支持改边 YdrMaster 2023-09-19 17:46:35 +0800
  • 7bdaea97ba fix: 改正导出中的笔误 YdrMaster 2023-09-19 17:10:16 +0800
  • 216b4e0356 feat: support communication op zhangyunze 2023-09-19 14:34:37 +0800
  • 742e876a96 fix: 改正一些算子的导出 YdrMaster 2023-09-19 14:17:55 +0800
  • fca88cc6ba feat: support some ops zhangyunze 2023-09-19 14:12:32 +0800
  • 8f2597a508
    fix bang runtime bug after merging distributed branch (#137) Haojie Wang 2023-09-19 14:10:39 +0800
  • da207b2638 Modify example file path to adapt dependency path change Baoming Linux Desktop 2023-09-18 22:27:30 -0700
  • 4fdcf97e32 fix: 正确应对匿名或重名的节点 YdrMaster 2023-09-19 10:54:01 +0800
  • ca149065d4 feat: 支持用 numpy.ndarray 构造张量 YdrMaster 2023-09-18 18:13:01 +0800
  • 8e4e392a49 feat: 导出数据 YdrMaster 2023-09-18 16:57:41 +0800
  • 04c2725312 feat: 正确传入数据并保存输出 YdrMaster 2023-09-18 16:50:00 +0800
  • 3f297b9757 Merge commit '48ec730579c7e5f5da203f4a77991cee5ccf8b82' into dev-dynamic-graph YdrMaster 2023-09-18 16:29:27 +0800
  • 030fdc0bd5 fmt YdrMaster 2023-09-18 16:25:42 +0800
  • a8f8d504f4 feat: support new graph to old runtime zhangyunze 2023-09-18 10:30:23 +0800
  • 48ec730579
    Support kvcache (#134) kilinchange 2023-09-18 14:17:02 +0800
  • e637f9e7dd feat: 支持导入和推导通信算子 YdrMaster 2023-09-18 12:07:25 +0800
  • 97439bd860 build: 删除无效代码 YdrMaster 2023-09-16 17:14:18 +0800
  • 24db46f01e feat: 支持导出所有张量的信息,只差数据了 YdrMaster 2023-09-16 16:36:56 +0800
  • d5a7713c7d refactor: 非 onnx 算子也能导出为自定义算子 YdrMaster 2023-09-16 14:24:29 +0800
  • a795364612 feat: 初步支持模型导出,不能导出属性张量,也不能导出张量的数据 YdrMaster 2023-09-16 14:20:00 +0800
  • c6b82cfda0
    Copyout numpy接口 (#135) PanZezhong1725 2023-09-15 16:40:44 +0800
  • 229993f081 fix: 前端需要载入所有连入图拓扑的边 YdrMaster 2023-09-15 10:27:40 +0800
  • eda206e2d0 build: Refactor Graph 切换分支 YdrMaster 2023-09-15 08:46:00 +0800
  • 39f1310343 fix format panzezhong 2023-09-14 16:47:47 +0800
  • 6ce2111842 Add dump interface panzezhong 2023-09-14 16:35:31 +0800
  • 4c321c8a91
    tensor parallel for transformer (#125) constroy Li 2023-09-14 14:19:45 +0800
  • 20a0c07992 feat: 支持值传入图 YdrMaster 2023-09-14 14:15:58 +0800
  • dda668fd16
    "modified where" (#131) xgqdut2016 2023-09-14 10:45:57 +0800
  • 6830f02c88 refactor: 直接导入 numpy ndarray 以提高加载性能 YdrMaster 2023-09-14 08:15:22 +0800
  • 5d56206d1a feat: 更新 RefactorGraph 以支持保存节点和边的名字 YdrMaster 2023-09-13 17:55:57 +0800
  • 18b844912d temp: update RefactorGraph YdrMaster 2023-09-13 14:34:16 +0800
  • 8eeccad5b3 build: update deps YdrMaster 2023-09-13 09:03:56 +0800
  • a6ad7b82f0 fix: 改正推导,可以开始折叠了 YdrMaster 2023-09-12 14:20:25 +0800
  • fd2d6b1926 feat: 支持代入值到变量 YdrMaster 2023-09-12 13:14:05 +0800
  • fd60190b1a fix: 成功进行一次形状推导 YdrMaster 2023-09-12 11:07:16 +0800
  • ffdddf3fec build: 不测 onnx 以过 ci YdrMaster 2023-09-11 17:47:05 +0800
  • 2efb337921 build: 更新 RefactorGraph 以支持算子注册 YdrMaster 2023-09-11 17:41:17 +0800
  • e57cdb5cec style: 产出没用的 ffi YdrMaster 2023-09-11 14:31:52 +0800
  • ea1d2c0129 feat: 支持 tensor attribute YdrMaster 2023-09-08 16:57:38 +0800
  • ed66ab6a37 feat: 导出新前端的 3 个函数并修改 python YdrMaster 2023-09-08 14:19:58 +0800
  • a8c117a48c build: 添加 RefactorGraph 子模块 YdrMaster 2023-09-08 11:06:24 +0800
  • bb20e55575 Merge branch 'dynamic_tensor' of github.com:InfiniTensor/InfiniTensor into dynamic_tensor kilinchange 2023-09-06 16:02:48 +0800
  • 74b382fd95 - fix allocator init kilinchange 2023-09-06 15:57:18 +0800
  • 836226448e fix bugs of creating multiple opt model stubs Xiaonan Song 2023-09-06 14:59:55 +0800
  • 08e271584a Merge branch 'dynamic_tensor' of github.com:InfiniTensor/InfiniTensor into dynamic_tensor kilinchange 2023-09-06 11:37:59 +0800
  • 648ed46952 - add tensorType - modify allocator to support memory allocation based on tensorType kilinchange 2023-09-06 11:28:13 +0800
  • 8c985bf752 fix bug introduced while resolving conflict Haojie Wang 2023-09-06 01:08:01 +0000
  • 436e7ede60
    Merge branch 'master' into xpu Haojie Wang 2023-09-06 08:55:55 +0800
  • c38d26bee8 fix: infershape should change some op private params zhangyunze 2023-09-05 11:30:22 +0800
  • 5375d529db - modify allocator kilinchange 2023-09-04 19:20:35 +0800
  • 04ba4b2d4b fix: fix the reshape op; fix the infershape to no const zhangyunze 2023-09-01 14:45:51 +0800
  • 02d02b1129 'feat: support dynamic input tensor zhangyunze 2023-08-31 15:40:00 +0800
  • f60767a770
    impl distributed launch with NCCL (#106) constroy Li 2023-09-05 09:47:35 +0800
  • e18900128d - modify allocator kilinchange 2023-09-04 19:20:35 +0800
  • b4eda85e67
    Fix mlu (#87) Hardy 2023-09-04 08:33:28 +0800
  • f000e211f7 fix: fix the reshape op; fix the infershape to no const zhangyunze 2023-09-01 14:45:51 +0800
  • 973f8a9030 'feat: support dynamic input tensor zhangyunze 2023-08-31 15:40:00 +0800
  • 2412c25e67
    Issue 107: Add copyin Numpy and covertion to Numpy (#126) PanZezhong1725 2023-09-01 11:20:26 +0800
  • 879f2f0315
    Merge branch 'master' into xpu Haojie Wang 2023-08-30 09:20:51 +0800
  • 3e6ef305f1
    框架支持bert/gpt2模型构图 (#94) gpt zhangyunze 2023-08-29 16:06:52 +0800
  • d8ffd8a4b7
    feat(env): add docker support. (#122) ChengXiang Qi 2023-08-28 18:34:36 +0800
  • a8a5c037ca feat(env): add docker support. kuangjux 2023-08-26 21:02:57 +0800
  • 85b96d8997
    Xpu abs (#121) zhangyue207 2023-08-28 10:15:07 +0800
  • 69fd251e5d
    Fix kernel arguments, add debug mode (#119) PanZezhong1725 2023-08-28 08:58:38 +0800
  • a4c6214529
    添加XPU算子 (#120) Bolun Zhang 2023-08-24 16:47:50 +0800
  • 0ce7e7651f Fix kernel arguments, add debug mode panzezhong 2023-08-24 13:39:22 +0800
  • 1e91979c76
    add CUDNN impl for Min and Max (#118) constroy Li 2023-08-22 16:19:29 +0800
  • d18d40a2e9 fix xpu code, let it can run. wanghailu 2023-08-22 15:42:53 +0800
  • 1438f14a25
    fix: fix castType mlu (#117) zhangyunze 2023-08-22 14:54:32 +0800
  • 9cf6c30e1c
    Add cuda transpose kernel (#115) PanZezhong1725 2023-08-22 14:22:15 +0800
  • 384407421b
    cudnn activations support ND-Tensor (#116) constroy Li 2023-08-22 14:21:59 +0800
  • a69390e310 Merge branch 'master' into xpu wanghailu 2023-08-22 09:17:23 +0800
  • 48847958d0
    impl sqrt on CUDA (#109) nnet_e2e_for_merge constroy Li 2023-08-18 12:17:47 +0800
  • ef672894d0
    support mixed dtype (#102) zhangyunze 2023-08-16 21:49:43 +0800
  • 0dc5347089
    memory_allocator (#103) kilinchange 2023-08-13 13:39:35 +0800
  • bd9e1aeb3f
    fix: fix cuda conv_fp16 run fail (#105) zhangyunze 2023-08-10 15:22:18 +0800
  • 8737f5773d commit for save gencode wanghailu 2023-08-10 15:09:19 +0800
  • fbaa5116b4 commit for save wanghailu 2023-08-08 16:36:37 +0800
  • 5472c3c61b add code wanghailu 2023-08-08 14:16:05 +0800
  • 4eb55542de commit for save wanghailu 2023-08-08 11:41:57 +0800
  • a7972bbb9b commit for save wanghailu 2023-08-08 10:43:57 +0800
  • 0699ffbd13 commit for save wanghailu 2023-08-07 17:26:16 +0800
  • 57ac94d893
    refactor(core): 添加新的 `OpType` 定义 (#99) Derui Yang 2023-08-07 11:17:05 +0800
  • c077a61681 fix: 改正 1x1 卷积的变换 optimization-pass YdrMaster 2023-04-26 12:20:17 +0800
  • 325b279468 typo YdrMaster 2023-04-25 17:33:54 +0800
  • 7b23fdbbfe feat: 基本实现空洞卷积的变换 YdrMaster 2023-04-25 15:59:43 +0800
  • e3428d8fd8 fix: 正确使用张量中的数据 YdrMaster 2023-04-24 13:56:04 +0800
  • 3631b03e73 feat: 优化转置写法 YdrMaster 2023-04-24 10:41:57 +0800
  • f78ff0e8ee feat: 实现 1x1 卷积转矩阵乘 YdrMaster 2023-04-23 18:14:51 +0800
  • 4a5e66b36a feat: 区分 1x1 卷积和空洞卷积 YdrMaster 2023-04-23 16:52:24 +0800
  • 7d7d923e8d perf: 没有变体就不用评分 YdrMaster 2023-04-23 16:22:47 +0800
  • 2a147c235d style: 调整结构 YdrMaster 2023-04-23 16:02:38 +0800
  • 17033fad97 feat: 添加导出 YdrMaster 2023-04-23 13:47:02 +0800
  • 28287f3782 feat: 完成导入 YdrMaster 2023-04-23 13:01:42 +0800