* feat!: Lock bazel version ([25f4371](https://github.com/NVIDIA/TRTorch/commit/25f4371))
* refactor(//cpp/api)!: Refactoring ptq to use includes but seperate from ([d2f8a59](https://github.com/NVIDIA/TRTorch/commit/d2f8a59))
Bug Fixes
* **//core:** Do not compile hidden methods ([6bd1a3f](https://github.com/NVIDIA/TRTorch/commit/6bd1a3f))
* **//core/conversion:** Check for calibrator before setting int8 mode ([3afd209](https://github.com/NVIDIA/TRTorch/commit/3afd209))
* **//core/conversion:** Supress unnecessary debug messages ([2b23874](https://github.com/NVIDIA/TRTorch/commit/2b23874))
* **//core/conversion/conversionctx:** Check both tensor and eval maps ([2d65ece](https://github.com/NVIDIA/TRTorch/commit/2d65ece))
* **//core/conversion/conversionctx:** In the case of strict types and ([3611778](https://github.com/NVIDIA/TRTorch/commit/3611778))
* **//core/conversion/converters:** Fix plugin implementation for TRT 7 ([94d6a0f](https://github.com/NVIDIA/TRTorch/commit/94d6a0f))
* **//core/conversion/converters/impl:** 1d case not working ([f42562b](https://github.com/NVIDIA/TRTorch/commit/f42562b))
* **//core/conversion/converters/impl:** code works for interpolate2d/3d, doesn't work for 1d yet ([e4cb117](https://github.com/NVIDIA/TRTorch/commit/e4cb117))
* **//core/conversion/converters/impl:** Fix interpolate.cpp ([b6942a2](https://github.com/NVIDIA/TRTorch/commit/b6942a2))
* **//core/conversion/converters/impl/element_wise:** Fix broadcast ([a9f33e4](https://github.com/NVIDIA/TRTorch/commit/a9f33e4))
* **//core/conversion/evaluators:** A couple fixes for evaluators ([07ba980](https://github.com/NVIDIA/TRTorch/commit/07ba980))
* **//core/lowering:** Conv2D -> _convolution pass was triggering conv ([ca2b5f9](https://github.com/NVIDIA/TRTorch/commit/ca2b5f9))
* **//cpp:** Remove deprecated script namespace ([d70760f](https://github.com/NVIDIA/TRTorch/commit/d70760f))
* **//cpp/api:** Better inital condition for the dataloader iterator to ([8d22bdd](https://github.com/NVIDIA/TRTorch/commit/8d22bdd))
* **//cpp/api:** Remove unecessary destructor in ptq class ([fc70267](https://github.com/NVIDIA/TRTorch/commit/fc70267))
* **//cpp/api:** set a default for calibrator ([825be69](https://github.com/NVIDIA/TRTorch/commit/825be69))
* **//cpp/benchmark:** reorder benchmark so FP16 bn issue in JIT doesnt ([98527d2](https://github.com/NVIDIA/TRTorch/commit/98527d2))
* **//cpp/ptq:** Default version of the app should not resize images ([de3cbc4](https://github.com/NVIDIA/TRTorch/commit/de3cbc4))
* **//cpp/ptq:** Enable FP16 kernels for INT8 applications ([26709cc](https://github.com/NVIDIA/TRTorch/commit/26709cc))
* **//cpp/ptq:** Enable FP16 kernels for INT8 applications ([e1c5416](https://github.com/NVIDIA/TRTorch/commit/e1c5416))
* **//cpp/ptq:** remove some logging from ptq app ([b989c7f](https://github.com/NVIDIA/TRTorch/commit/b989c7f))
* **//cpp/ptq:** Tracing model in eval mode wrecks accuracy in Libtorch ([54a24b3](https://github.com/NVIDIA/TRTorch/commit/54a24b3))
* **//cpp/trtorchc:** Refactor trtorchc to use new C++ API ([789e1be](https://github.com/NVIDIA/TRTorch/commit/789e1be)), closes [#132](https://github.com/NVIDIA/TRTorch/issues/132)
* **//cpp/trtorchc:** Support building trtorchc with the pre_cxx11_abi ([172d4d5](https://github.com/NVIDIA/TRTorch/commit/172d4d5))
* **//docs:** add nojekyll file ([2a02cd5](https://github.com/NVIDIA/TRTorch/commit/2a02cd5))
* **//docs:** fix version links ([11555f7](https://github.com/NVIDIA/TRTorch/commit/11555f7))
* **//notebooks:** Fix WORKSPACE template file to reflect new build system layout ([c8ea9b7](https://github.com/NVIDIA/TRTorch/commit/c8ea9b7))
* **//py:** Build system issues ([c1de126](https://github.com/NVIDIA/TRTorch/commit/c1de126))
* **//py:** Ignore generated version file ([9e37dc1](https://github.com/NVIDIA/TRTorch/commit/9e37dc1))
* **//py:** Lib path incorrect ([ff2b13c](https://github.com/NVIDIA/TRTorch/commit/ff2b13c))
* **//tests:** Duplicated tensorrt dep ([5cd697e](https://github.com/NVIDIA/TRTorch/commit/5cd697e))
* **//third_party/tensorrt:** Fix include dir for library headers ([22ed5cf](https://github.com/NVIDIA/TRTorch/commit/22ed5cf))
* **//third_party/tensorrt:** Fix TensorRT paths for local x86 builds ([73d804b](https://github.com/NVIDIA/TRTorch/commit/73d804b))
* **aarch64:** fixes and issues for aarch64 toolchain ([9a6cccd](https://github.com/NVIDIA/TRTorch/commit/9a6cccd))
* **aten::_convolution:** out channels was passed in incorrectly for ([ee727f8](https://github.com/NVIDIA/TRTorch/commit/ee727f8))
* **aten::_convolution:** Pass dummy bias when there is no bias ([b20671c](https://github.com/NVIDIA/TRTorch/commit/b20671c))
* **aten::batch_norm:** A new batch norm implementation that hopefully ([6461872](https://github.com/NVIDIA/TRTorch/commit/6461872))
* **aten::batchnorm|aten::view:** Fix converter implementation for ([bf651dd](https://github.com/NVIDIA/TRTorch/commit/bf651dd))
* **aten::contiguous:** Blacklist aten::contiguous from conversion ([b718121](https://github.com/NVIDIA/TRTorch/commit/b718121))
* **aten::flatten:** Fixes dynamic shape for flatten ([4eb20bb](https://github.com/NVIDIA/TRTorch/commit/4eb20bb))
* fixed FP16 bug, fixed README, addressed some other PR comments ([d9c0e84](https://github.com/NVIDIA/TRTorch/commit/d9c0e84))
* **aten::neg:** Fix a index bug in neg ([1b2cde4](https://github.com/NVIDIA/TRTorch/commit/1b2cde4))
* **aten::size, other aten evaluators:** Removes aten::size converter in ([c83447e](https://github.com/NVIDIA/TRTorch/commit/c83447e))
* **BUILD:** modified BUILD ([a0d8586](https://github.com/NVIDIA/TRTorch/commit/a0d8586))
* trying to resolve interpolate plugin problems ([f0fefaa](https://github.com/NVIDIA/TRTorch/commit/f0fefaa))
* **core/conversion/converters/impl:** fix error message in interpolate ([5ddab8b](https://github.com/NVIDIA/TRTorch/commit/5ddab8b))
* Address issues in PR ([cd24f26](https://github.com/NVIDIA/TRTorch/commit/cd24f26))
* bypass jeykll, also add PR template ([a41c400](https://github.com/NVIDIA/TRTorch/commit/a41c400))
* first commit ([4f1a9df](https://github.com/NVIDIA/TRTorch/commit/4f1a9df))
* Fix pre CXX11 ABI python builds and regen docs ([42013ab](https://github.com/NVIDIA/TRTorch/commit/42013ab))
* fixed interpolate_plugin to handle dynamically sized inputs for adaptive_pool2d ([7794c78](https://github.com/NVIDIA/TRTorch/commit/7794c78))
* need to fix gather converter ([024a6b2](https://github.com/NVIDIA/TRTorch/commit/024a6b2))
* **plugin:** trying to fix bug in plugin ([cafcced](https://github.com/NVIDIA/TRTorch/commit/cafcced))
* **pooling:** fix the tests and the 1D pooling cases ([a90e6db](https://github.com/NVIDIA/TRTorch/commit/a90e6db))
* RunGraphEngineDynamic fixed to work with dynamically sized input tensors ([6308190](https://github.com/NVIDIA/TRTorch/commit/6308190))
Features
* **//:libtrtorch:** Ship trtorchc with the tarball ([d647447](https://github.com/NVIDIA/TRTorch/commit/d647447))
* **//core/compiler:** Multiple outputs supported now via tuple ([f9af574](https://github.com/NVIDIA/TRTorch/commit/f9af574))
* **//core/conversion:** Adds the ability to evaluate loops ([dcb1474](https://github.com/NVIDIA/TRTorch/commit/dcb1474))
* **//core/conversion:** Compiler can now create graphs ([9d1946e](https://github.com/NVIDIA/TRTorch/commit/9d1946e))
* **//core/conversion:** Evaluation of static conditionals works now ([6421f3d](https://github.com/NVIDIA/TRTorch/commit/6421f3d))
* **//core/conversion/conversionctx:** Make op precision available at ([78a1c61](https://github.com/NVIDIA/TRTorch/commit/78a1c61))
* **//core/conversion/converters:** Throw a warning if a converter is ([6cce381](https://github.com/NVIDIA/TRTorch/commit/6cce381))
* **//core/conversion/converters/impl:** added support for aten::stack ([415378e](https://github.com/NVIDIA/TRTorch/commit/415378e))
* **//core/conversion/converters/impl:** added support for linear1d and bilinear2d ops ([4416d1f](https://github.com/NVIDIA/TRTorch/commit/4416d1f))
* **//core/conversion/converters/impl:** added support for trilinear3d op ([bb46e70](https://github.com/NVIDIA/TRTorch/commit/bb46e70))
* **//core/conversion/converters/impl:** all function schemas for upsample_nearest ([1b50484](https://github.com/NVIDIA/TRTorch/commit/1b50484))
* **//core/conversion/converters/impl:** logic implemented ([7f12160](https://github.com/NVIDIA/TRTorch/commit/7f12160))
* **//core/conversion/converters/impl:** Round out pooling ([7dc4af4](https://github.com/NVIDIA/TRTorch/commit/7dc4af4))
* **//core/conversion/converters/impl:** select converter, which adds support for aten::select.int ([5151c34](https://github.com/NVIDIA/TRTorch/commit/5151c34))
* **//core/conversion/converters/impl/plugins:** Created interpolate plugin, works for mode='linear' ([205ab99](https://github.com/NVIDIA/TRTorch/commit/205ab99))
* **//core/conversion/converters/impl/plugins:** interpolate plugin compiles now. time to test it. ([58dbaef](https://github.com/NVIDIA/TRTorch/commit/58dbaef))
* **//core/conversion/converters/impl/plugins:** template for interpolate plugin ([7c91dec](https://github.com/NVIDIA/TRTorch/commit/7c91dec))
* **//core/conversion/converters/impl/shuffle:** Implement aten::resize ([353f2d2](https://github.com/NVIDIA/TRTorch/commit/353f2d2))
* **//core/conversion/evaluators:** A whole bunch of new evaluators ([7466b8a](https://github.com/NVIDIA/TRTorch/commit/7466b8a))
* **//core/conversion/evaluators:** adding support for common evaluation ([d351717](https://github.com/NVIDIA/TRTorch/commit/d351717))
* **//core/conversion/evaluators:** Adds new applicability filters for ([2cc3226](https://github.com/NVIDIA/TRTorch/commit/2cc3226))
* **//core/conversion/evaluators:** Allow ITensors to be wrapped in ([619e345](https://github.com/NVIDIA/TRTorch/commit/619e345))
* **//core/execution:** Type checking for the executor, now is the ([2dd1ba3](https://github.com/NVIDIA/TRTorch/commit/2dd1ba3))
* **//core/lowering:** Add tuple lowering pass to remove tuples if ([ce6cf75](https://github.com/NVIDIA/TRTorch/commit/ce6cf75))
* **//core/lowering:** Adds peephole optimization pass ([0014b84](https://github.com/NVIDIA/TRTorch/commit/0014b84))
* **//core/lowering:** Fuse aten::addmm branches into a single ([68f0317](https://github.com/NVIDIA/TRTorch/commit/68f0317))
* **//core/lowering:** New freeze model pass and new exception ([4acc3fd](https://github.com/NVIDIA/TRTorch/commit/4acc3fd))
* **//core/lowering:** Remove aten::contiguous ([630b615](https://github.com/NVIDIA/TRTorch/commit/630b615))
* **//core/quantization:** skeleton of INT8 PTQ calibrator ([dd443a6](https://github.com/NVIDIA/TRTorch/commit/dd443a6))
* **//core/util:** New logging level for Graph Dumping ([90c44b9](https://github.com/NVIDIA/TRTorch/commit/90c44b9))
* **//cpp/api:** Adding max batch size setting ([1b25542](https://github.com/NVIDIA/TRTorch/commit/1b25542))
* **//cpp/api:** Functional Dataloader based PTQ ([f022dfe](https://github.com/NVIDIA/TRTorch/commit/f022dfe))
* **//cpp/api:** Remove the extra includes in the API header ([2f86f84](https://github.com/NVIDIA/TRTorch/commit/2f86f84))
* **//cpp/benchmark:** Increased workspace size for benchmark, may help ([8171f79](https://github.com/NVIDIA/TRTorch/commit/8171f79))
* **//cpp/ptq:** Add a feature to the dataset to use less than the full ([5f36f47](https://github.com/NVIDIA/TRTorch/commit/5f36f47))
* **//cpp/ptq:** do real benchmarking in the PTQ app instead of rough ([65e71c7](https://github.com/NVIDIA/TRTorch/commit/65e71c7))
* **//cpp/ptq/training:** Training recipe for VGG16 Classifier on ([676bf56](https://github.com/NVIDIA/TRTorch/commit/676bf56))
* **//cpp/trtorchc:** Adding a new CLI application for TRTorch which ([4f349a1](https://github.com/NVIDIA/TRTorch/commit/4f349a1))
* **//cpp/trtorchexec:** TRTorch exec now supports checking correctness ([80808b7](https://github.com/NVIDIA/TRTorch/commit/80808b7))
* **//lowering:** centralize lowering and try to use PyTorch Conv2DBN folding ([fad4a10](https://github.com/NVIDIA/TRTorch/commit/fad4a10))
* **//py:** add the option to build python package with CXX11 abi ([fdbd7d2](https://github.com/NVIDIA/TRTorch/commit/fdbd7d2))
* **//py:** API now produces valid engines that are consumable by ([72bc1f7](https://github.com/NVIDIA/TRTorch/commit/72bc1f7))
* **//py:** Inital introduction of the Python API ([7088245](https://github.com/NVIDIA/TRTorch/commit/7088245))
* **//py:** Manylinux container and build system for multiple python ([639c2a3](https://github.com/NVIDIA/TRTorch/commit/639c2a3))
* **//py:** register trtorch with torch op library to support ([736e914](https://github.com/NVIDIA/TRTorch/commit/736e914))
* **//py:** setup.py now searches for bazel executable ([737fe5c](https://github.com/NVIDIA/TRTorch/commit/737fe5c))
* **//py:** Working portable package ([482ef2c](https://github.com/NVIDIA/TRTorch/commit/482ef2c))
* added adaptive_avg_pool2d plugin, and added test for it ([fa227b0](https://github.com/NVIDIA/TRTorch/commit/fa227b0))
* **//tests:** New optional accuracy tests to check INT8 and FP16 ([df74136](https://github.com/NVIDIA/TRTorch/commit/df74136))
* **//toolchains:** Adding platform targets for supported platforms ([7889ebd](https://github.com/NVIDIA/TRTorch/commit/7889ebd))
* **/cpp/api:** Working INT8 Calibrator, also resolves [41](https://github.com/NVIDIA/TRTorch/issues/41) ([5c0d737](https://github.com/NVIDIA/TRTorch/commit/5c0d737))
* **aten::add_t:** aten::add_.t evaluator that adds lists together ([c4c3ce1](https://github.com/NVIDIA/TRTorch/commit/c4c3ce1))
* **aten::avg_pool2d:** Implement Average Pooling 2D ([0c39519](https://github.com/NVIDIA/TRTorch/commit/0c39519))
* **aten::cat:** Implements aten::cat and completes support for SSD ([c2d3a6e](https://github.com/NVIDIA/TRTorch/commit/c2d3a6e))
* **aten::conv_transpose:** Add support for dilated and group ([48b950a](https://github.com/NVIDIA/TRTorch/commit/48b950a))
* **aten::dropout_:** Remove inplace dropout ([7aa57c3](https://github.com/NVIDIA/TRTorch/commit/7aa57c3))
* **aten::flatten:** Adds a converter for aten flatten since MM is the ([d945eb9](https://github.com/NVIDIA/TRTorch/commit/d945eb9))
* addressed some PR comments, refactored code ([141763f](https://github.com/NVIDIA/TRTorch/commit/141763f))
* **aten::matmul|aten::addmm:** Adds support for aten::matmul and ([c5b6202](https://github.com/NVIDIA/TRTorch/commit/c5b6202))
* **aten::permute:** Implement permute support ([c7d6b49](https://github.com/NVIDIA/TRTorch/commit/c7d6b49))
* **aten::size [static]:** Implement a aten::size converter for static input size ([0548540](https://github.com/NVIDIA/TRTorch/commit/0548540))
* started to work on add_.t evaluator, doesn't work yet ([f216d3f](https://github.com/NVIDIA/TRTorch/commit/f216d3f))
* **aten::to:** Remove remaining typecast operators (should be a very ([0f63ffa](https://github.com/NVIDIA/TRTorch/commit/0f63ffa))
* **aten::view:** Adds support for ATen view also fixes some tests ([24b422e](https://github.com/NVIDIA/TRTorch/commit/24b422e))
* **aten::zeros:** Implement aten::zeros evaluator ([670817c](https://github.com/NVIDIA/TRTorch/commit/670817c))
* **conv2d_to_convolution:** A pass to map aten::conv2d to _convolution ([2c5c0d5](https://github.com/NVIDIA/TRTorch/commit/2c5c0d5))
* **prim::NumToTensor:** Implement evaluator for NumToTensor ([60df888](https://github.com/NVIDIA/TRTorch/commit/60df888))
* **tests/util:** added RunGraphEngineDynamic to handle dynamic input sized tensors ([9458f21](https://github.com/NVIDIA/TRTorch/commit/9458f21))
* **trt_util:** from Naren, added unpadDims tool ([164a1a6](https://github.com/NVIDIA/TRTorch/commit/164a1a6))
* support for adaptive_avg_pool2d plugin ([52be580](https://github.com/NVIDIA/TRTorch/commit/52be580))
* Support non cxx11-abi builds for use in python api ([83e0ed6](https://github.com/NVIDIA/TRTorch/commit/83e0ed6))
BREAKING CHANGES
* Bazel version is now locked to Bazel 3.3.1 and will be
bumped manually from now on. Builds will fail on all other versions
since now bazel will check the version before it compiles.
Documentation on how to install bazel is added as well to support
aarch64 until bazel releases binaries for the platform (which is soon)
Signed-off-by: Naren Dasan <narennarendasan.com>
Signed-off-by: Naren Dasan <narensnvidia.com>
* To use ptq you now need to include trtorch/ptq.h in
addition to trtorch/trtorch.h, similarly for logging commands you need
to include trtorch/logging.h
Signed-off-by: Naren Dasan <narennarendasan.com>
Signed-off-by: Naren Dasan <narensnvidia.com>