github-featured/nvidia--tensorrt
2
0
Fork 0
Code Issues Pull Requests Deployments Models Agent Notes Packages Projects Releases Wiki Activity

Commit Graph

  • fd852d73d1 Add benchmark data for demoBERT-MT w/ sparsity and int8-qat Rajeev Rao 2021-08-04 10:22:08 -07:00
  • 834cb6d05b Update make command for native aarch64 build Rajeev Rao 2021-07-12 02:28:23 -07:00
  • 2a0107cc59 Remove anchors for V100 benchmark results in demoBERT README. Rajeev Rao 2021-07-09 16:28:53 +00:00
  • 3428f1173d Analytics tracker for ONNX-GS and Pytorch Quantization Toolkit Python docs Pranav Marathe 2021-07-09 10:15:03 -04:00
  • ef01e58a25 EfficientNMS Updates: Empty Input Tensor + TX1/TX2 Resources William Raveane 2021-07-01 18:07:23 -04:00
  • 4234b20fad Mark ONNX NMS operator as experimental Rajeev Rao 2021-07-08 13:56:00 -07:00
  • 6198e4c5c1 Fix typo in BERT notebook Rajeev Rao 2021-07-08 11:22:57 -07:00
  • 56652dd686 Update demoBERT benchmark data from PBR# 156504 Rajeev Rao 2021-07-08 17:07:25 +00:00
  • b325b64871 Clear outputs in demoBERT notebooks Rajeev Rao 2021-07-07 14:36:36 -07:00
  • 43e859e269 Add EfficientDet sample notebook Shashank Verma 2021-07-07 18:19:09 +00:00
  • 8b30eaeadb Add demoBERT notebooks Vinh Nguyen 2021-07-07 00:13:54 +00:00
  • c2668947ea Update make command for native aarch64 build Rajeev Rao 2021-07-12 02:28:23 -07:00
  • 60024121da Remove anchors for V100 benchmark results in demoBERT README. Rajeev Rao 2021-07-09 16:28:53 +00:00
  • 7e88742def Analytics tracker for ONNX-GS and Pytorch Quantization Toolkit Python docs Pranav Marathe 2021-07-09 10:15:03 -04:00
  • 3235cc2ffc EfficientNMS Updates: Empty Input Tensor + TX1/TX2 Resources William Raveane 2021-07-01 18:07:23 -04:00
  • ac59f5ffaf Mark ONNX NMS operator as experimental Rajeev Rao 2021-07-08 13:56:00 -07:00
  • eeca567366 Fix typo in BERT notebook 21.07 Rajeev Rao 2021-07-08 11:22:57 -07:00
  • 718a92f5a0 Update demoBERT benchmark data from PBR# 156504 Rajeev Rao 2021-07-08 17:07:25 +00:00
  • ddaf0b9e1b Clear outputs in demoBERT notebooks Rajeev Rao 2021-07-07 14:36:36 -07:00
  • 731629abe7 Add EfficientDet sample notebook Shashank Verma 2021-07-07 18:19:09 +00:00
  • 8c8344f8a9 Add demoBERT notebooks Vinh Nguyen 2021-07-07 00:13:54 +00:00
  • eb5de99b52 Update ONNX-TensorRT to v8.0 8.0.1 Rajeev Rao 2021-07-02 12:09:19 -07:00
  • aff45dd565 TensorRT OSS 8.0 release Rajeev Rao 2021-07-02 00:21:09 -07:00
  • d04182cd00 Revert "remove fcplugin to accelerate pipeline run time" Rajeev Rao 2021-06-23 13:37:20 -07:00
  • a2b3d3d5cc Add switch for batch agnostic mode in NMS plugin 21.06 Kevin Chen 2021-06-22 22:09:32 -07:00
  • 0953f2ff87 Update ONNX-GraphSurgeon to v0.3.9 Rajeev Rao 2021-06-22 15:29:02 -07:00
  • 10755383a4 Add 21.06 Changelog Rajeev Rao 2021-06-22 15:19:40 -07:00
  • 73fa0193de Update Polygraphy to v0.29.2 Rajeev Rao 2021-06-22 15:23:02 -07:00
  • 593684b279 Cleanup uff_custom_plugin sample Rajeev Rao 2021-06-22 15:19:08 -07:00
  • d4a5e08a3b Clang-tidy plugins Rajeev Rao 2021-06-22 15:18:27 -07:00
  • d742af876d Optimize TLT MaskRCNN plugins Tyler Zhu 2021-06-02 16:37:47 +08:00
  • c5b8cf67fb Fix numerical errors for floats in NMS/batchedNMS plugins. zhimengf@nvidia.com 2021-04-23 09:14:41 +08:00
  • b2d60b6e10 GHSA-cwv3-863g-39vx - do not specify tf1 in Polygraphy requirements file Rajeev Rao 2021-06-09 23:34:59 -07:00
  • b30bf6c2bb Quickstart update (#1171) dnola 2021-06-09 22:57:16 -07:00
  • 611dba6388 CVE-2021-27921 - update Pillow version to 8.1.2 21.05 Rajeev Rao 2021-05-19 14:04:48 -07:00
  • 8593e51b09 remove fcplugin to accelerate pipeline run time Vincent Huang 2021-04-16 10:15:16 +00:00
  • 19e3899932 speed up engine build for bert on TRT-8.0 Vincent Huang 2021-04-15 02:33:04 +00:00
  • f00665c362 Add InstanceNorm3D plugin and update ONNX parser to 21.05 Rajeev Rao 2021-05-19 10:48:16 -07:00
  • b855656005 TensorRT-OSS 21.05 release Rajeev Rao 2021-05-19 10:20:04 -07:00
  • ab20a8ac26 Update demoBERT input dimensions to match Triton requirement (#1051) Dong 2021-05-11 04:02:37 -07:00
  • c9c1327664 Add model.py and lenet5.py (#1232) Kevin Chen 2021-05-04 16:38:24 -07:00
  • dec26f02eb Add email for press contacts Rajeev Rao 2021-04-21 13:35:56 -07:00
  • 3835424af0 Populate QKV plugin attributes (#1190) Kevin Chen 2021-04-13 15:30:57 -07:00
  • 4c99d07241 TensorRT OSS 21.04 release - update Changelog 21.04 Rajeev Rao 2021-04-12 12:17:29 -07:00
  • 64fa61b415 Add varlen MHA fp16 slen=384 kernels for sm_86 Rajeev Rao 2021-04-12 11:53:49 -07:00
  • 7f087a72f0 Update ONNX parser Rajeev Rao 2021-04-12 11:52:16 -07:00
  • 2e875dbb28 fix doc for bert plugins Vincent Huang 2021-03-12 00:27:43 -08:00
  • c4dbccc694 Samples refresh and bugfixes Rajeev Rao 2021-03-17 00:37:41 -07:00
  • 80e2473bdb serialize CPU data of PriorBox to engine only Rajeev Rao 2021-03-18 03:35:09 -07:00
  • 932bcd6893 Avoid arithmetic on void pointer in multilevelProposeROIPlugin.cpp #1028 Rajeev Rao 2021-03-18 22:29:54 -07:00
  • 1fe5532c0c Allow MHA plugin to run on SM_86 as well Rajeev Rao 2021-03-18 01:57:07 -07:00
  • 853d3311e8 Update TensorRT headers to 7.2.3.4 Rajeev Rao 2021-03-16 23:07:39 -07:00
  • cc5dede848 clang-format plugins and samples Rajeev Rao 2021-03-16 23:05:01 -07:00
  • 8e52ae4060 Updates fold_constants to correctly exit early when there is nothing to fold Pranav Marathe 2021-03-10 15:34:18 -05:00
  • d3e65fc976 Permit groupadd with non-unique GID in build containers Rajeev Rao 2021-03-18 22:55:56 -07:00
  • 5094c1dea6 Update TensorRT versions to 7.2.3.4 Rajeev Rao 2021-03-16 23:07:54 -07:00
  • da405547b0 Removes extra terminate call in InstanceNorm Rajeev Rao 2021-03-18 03:18:01 -07:00
  • 3622eccca8 Fix broken link in sampleUffMaskRCNN README Rajeev Rao 2021-04-05 22:03:26 -07:00
  • e724d31ab8 Set CUDA_INSTALL_DIR if it is not set by cmake build command Nishidha Panpaliya 2021-03-19 07:46:32 -05:00
  • 5a486140e8 Fix bug UID 0 is not unique Dong Wang 2021-02-28 10:39:05 +01:00
  • 0dcade2184 don't violate strict aliasing with reinterpret_cast Tom Peters 2019-10-01 14:57:05 -04:00
  • 4e5e724891 Use right CMake variable for cu files in GroupNorm Mickaël Seznec 2021-02-24 16:30:53 +01:00
  • 89d7964215 Add onnx parser update to Changelog Rajeev Rao 2021-03-09 13:57:50 -08:00
  • 27cab2a710 Update onnx-graphsurgeon to v0.3.3 Rajeev Rao 2021-03-09 13:46:13 -08:00
  • e0818625af Update CHANGELOG for 21.03 Rajeev Rao 2021-03-09 09:11:34 -08:00
  • 19910a4c95 Bugfixes for decodeBBoxes kernel. zhimengf@nvidia.com 2021-02-18 16:10:55 +08:00
  • 068350b021 create ProposalDynamic and CropAndResizeDynamic plugins. zhimengf@nvidia.com 2021-02-02 13:24:47 +08:00
  • d5b878a18e optimized FP16 NMS/BatchedNMS plugins. zhimengf@nvidia.com 2021-02-01 13:54:19 +08:00
  • 10a51bed54 Fix a computational problem of scaledSoftmax. The original implementation results in wrong results of sum of softmax such that the results of BERT models (128 < seq_len < 384 and seq_len > 384) are very large or even 'nan'. This implementation fix the computational problem such that the results of BERT models (128 < seq_len < 384 and seq_len > 384) become correct. yuanzexi 2021-03-04 16:38:12 +08:00
  • c8a30a4557 Add onnx parser update to Changelog 21.03 Rajeev Rao 2021-03-09 13:57:50 -08:00
  • cebbd00a49 Update onnx-graphsurgeon to v0.3.3 Rajeev Rao 2021-03-09 13:46:13 -08:00
  • 7c015d0dc2 Update CHANGELOG for 21.03 Rajeev Rao 2021-03-09 09:11:34 -08:00
  • a69ba6a9b3 Bugfixes for decodeBBoxes kernel. zhimengf@nvidia.com 2021-02-18 16:10:55 +08:00
  • 7b93387207 create ProposalDynamic and CropAndResizeDynamic plugins. zhimengf@nvidia.com 2021-02-02 13:24:47 +08:00
  • 15128d0e02 optimized FP16 NMS/BatchedNMS plugins. zhimengf@nvidia.com 2021-02-01 13:54:19 +08:00
  • 8c10371601 remove redundant blank lines yuanzexi 2021-03-05 17:41:25 +08:00
  • cee2a97312 revert blank lines. yuanzexi 2021-03-05 16:38:41 +08:00
  • cc94e3eab6 add signed-off yuanzexi 2021-03-05 15:16:19 +08:00
  • dc519dd85c format the code yuanzexi 2021-03-05 15:10:11 +08:00
  • 27b8c50d43 Fix a computational problem of scaledSoftmax. The original implementation results in wrong results of sum of softmax such that the results of BERT models (128 < seq_len < 384 and seq_len > 384) are very large or even 'nan'. This implementation fix the computational problem such that the results of BERT models (128 < seq_len < 384 and seq_len > 384) become correct. yuanzexi 2021-03-04 16:38:12 +08:00
  • 88f9cae8af Merge pull request #1075 from NVIDIA/dev/pranavm/fix-onnx pranavm-nvidia 2021-02-22 13:20:34 -05:00
  • 45e8a8b7b7 Merge pull request #1076 from NVIDIA/onnx-gs-license pranavm-nvidia 2021-02-22 13:20:13 -05:00
  • 2b198694e8 Adds license file for ONNX-GraphSurgeon pranavm 2021-02-22 09:31:28 -08:00
  • ff1e15458c Fixes ONNX parser Python bindings pranavm 2021-02-22 09:30:35 -08:00
  • 4225306155 [Plugin] Fix bug of decodeBboxes in NMS plugin. zhimengf@nvidia.com 2021-02-19 13:43:23 +08:00
  • 13f7441a06 Fix function name mismatch error CryptoSalamander 2021-02-12 05:34:22 +09:00
  • 0a80821676 [FIX] Function Name Mismatch CryptoSalamander 2020-10-26 01:38:41 +09:00
  • d7baf010e4 TensorRT OSS 21.02 release 21.02 Rajeev Rao 2021-01-31 22:52:19 -08:00
  • 360eea3e14 Add monthly releases to CHANGELOG Rajeev Rao 2021-02-02 16:38:06 -08:00
  • 23adb1a30d Fix for triton dynamic batching requirement DougM 2020-11-08 22:39:41 -08:00
  • f23787eb00 Update VERSION Rajeev Rao 2021-01-27 17:50:19 -08:00
  • 183f891191 Update onnx-tensorrt and copyright headers (2021) Rajeev Rao 2021-01-13 23:22:30 -08:00
  • da988fd1f1 ONNX tools updates: Polygraphy, ONNX-GS Rajeev Rao 2021-01-13 23:21:56 -08:00
  • 0d538ddedc Explicit casts to resolve compiler errors in __half type conversion Rajeev Rao 2021-01-13 14:53:53 -08:00
  • aa1b4b09d6 fix failing NMS on fp16 due to issue with box-size calculation Paul Bridger 2020-12-24 07:04:02 +01:00
  • 2c942e3fb2 fix serialization size Paul Bridger 2020-12-21 02:30:01 +01:00
  • 42dbbb0ef3 add fp16 capability to batchedNMSPlugins Paul Bridger 2020-12-20 13:02:05 +01:00
  • 7ca28ecf1c Explicit casts to resolve compiler errors in __half type conversion Rajeev Rao 2021-01-13 14:53:53 -08:00
  • aa77506a90 fix failing NMS on fp16 due to issue with box-size calculation Paul Bridger 2020-12-24 07:04:02 +01:00
  • de72d29064 fix serialization size Paul Bridger 2020-12-21 02:30:01 +01:00