TRT 8.0 fixes for trtorchc

Signed-off-by: Boris Fomitchev <[email protected]>

Author: borisfom

WORKSPACE

@@ -34,8 +34,8 @@ git_repository(
 # CUDA should be installed on the system locally
 new_local_repository(
     name = "cuda",
-    path = "/usr/local/cuda",
     build_file = "@//third_party/cuda:BUILD",
+    path = "/usr/local/cuda-11.1/",
 )
 
 new_local_repository(
@@ -67,23 +67,25 @@ http_archive(
 # Either place them in the distdir directory in third_party and use the --distdir flag
 # or modify the urls to "file:///<PATH TO TARBALL>/<TARBALL NAME>.tar.gz
 
-#http_archive(
-#    name = "cudnn",
-#    urls = ["file:///home/boris/git/TRTorch/third_party/dist_dir/x86_64-linux-gnu/cudnn-11.2-linux-x64-v8.1.1.33.tgz",],
-#    # "https://developer.nvidia.com/compute/machine-learning/cudnn/secure/8.1.1.33/11.2_20210301/cudnn-11.2-linux-x64-v8.1.1.33.tgz",],
-#    build_file = "@//third_party/cudnn/archive:BUILD",
-#    sha256 = "98a8784e92862f20018d20c281b30d4a0cd951f93694f6433ccf4ae9c502ba6a",
-#    strip_prefix = "cuda"
-#)
+http_archive(
+    name = "cudnn",
+    build_file = "@//third_party/cudnn/archive:BUILD",
+    sha256 = "98a8784e92862f20018d20c281b30d4a0cd951f93694f6433ccf4ae9c502ba6a",
+    strip_prefix = "cuda",
+    urls = [
+        "https://developer.nvidia.com/compute/machine-learning/cudnn/secure/8.1.1.33/11.2_20210301/cudnn-11.2-linux-x64-v8.1.1.33.tgz",
+    ],
+)
 
-#http_archive(
-#    name = "tensorrt",
-#    urls = ["file:///home/boris/git/TRTorch/third_party/dist_dir/x86_64-linux-gnu/TensorRT-7.2.3.4.Ubuntu-18.04.x86_64-gnu.cuda-11.1.cudnn8.1.tar.gz"],
-# "https://developer.nvidia.com/compute/machine-learning/tensorrt/secure/7.2.3/tars/TensorRT-7.2.3.4.Ubuntu-18.04.x86_64-gnu.cuda-11.1.cudnn8.1.tar.gz",],
-#    build_file = "@//third_party/tensorrt/archive:BUILD",
-#    strip_prefix = "TensorRT-7.2.3.4",
-#    sha256 = "d3a1f478e304b48878604fac70ce7920fece71f9cac62f925c9c59c197f5d087"
-#)
+http_archive(
+    name = "tensorrt",
+    build_file = "@//third_party/tensorrt/archive:BUILD",
+    sha256 = "d3a1f478e304b48878604fac70ce7920fece71f9cac62f925c9c59c197f5d087",
+    strip_prefix = "TensorRT-7.2.3.4",
+    urls = [
+        "https://developer.nvidia.com/compute/machine-learning/tensorrt/secure/7.2.3/tars/TensorRT-7.2.3.4.Ubuntu-18.04.x86_64-gnu.cuda-11.1.cudnn8.1.tar.gz",
+    ],
+)
 
 ####################################################################################
 # Locally installed dependencies (use in cases of custom dependencies or aarch64)
@@ -108,17 +110,17 @@ http_archive(
 #    build_file = "third_party/libtorch/BUILD"
 #)
 
-new_local_repository(
-    name = "cudnn",
-    path = "/usr/",
-    build_file = "@//third_party/cudnn/local:BUILD"
-)
+#new_local_repository(
+#    name = "cudnn",
+#    path = "/usr/",
+#    build_file = "@//third_party/cudnn/local:BUILD"
+#)
 
-new_local_repository(
-   name = "tensorrt",
-   path = "/usr/",
-   build_file = "@//third_party/tensorrt/local:BUILD"
-)
+#new_local_repository(
+#   name = "tensorrt",
+#   path = "/usr/",
+#   build_file = "@//third_party/tensorrt/local:BUILD"
+#)
 
 #########################################################################
 # Testing Dependencies (optional - comment out on aarch64)

cpp/api/include/trtorch/ptq.h

@@ -77,7 +77,7 @@ class Int8Calibrator : Algorithm {
    *
    * @return int
    */
-  int getBatchSize() const override {
+  int getBatchSize() const noexcept override {
     // HACK: TRTorch only uses explict batch sizing, INT8 Calibrator does not
     // work when reporting the batch size here and having explicity batching.
     // So we just report batch size 1 (warnings will still be printed out).
@@ -96,7 +96,7 @@ class Int8Calibrator : Algorithm {
    * @return true - There is a new batch for the calibrator to consume
    * @return false - There is not a new batch for the calibrator to consume
    */
-  bool getBatch(void* bindings[], const char* names[], int nbBindings) override {
+  bool getBatch(void* bindings[], const char* names[], int nbBindings) noexcept override {
     if (it_ != batched_data_.end()) {
       auto status = get_batch_impl(bindings, names, nbBindings, *it_);
       it_ = ++it_;
@@ -116,7 +116,7 @@ class Int8Calibrator : Algorithm {
    * @param length
    * @return const void* - Pointer to cache data
    */
-  const void* readCalibrationCache(size_t& length) override {
+  const void* readCalibrationCache(size_t& length) noexcept override {
     if (use_cache_) {
       std::stringstream ss;
       ss << "Reading Calibration Cache from " << cache_file_path_;
@@ -143,7 +143,7 @@ class Int8Calibrator : Algorithm {
    * @param cache: const void* - cache data
    * @param length: size_t - length of cache
    */
-  void writeCalibrationCache(const void* cache, size_t length) override {
+  void writeCalibrationCache(const void* cache, size_t length) noexcept override {
     std::ofstream cache_file(cache_file_path_, std::ios::binary);
     cache_file.write(reinterpret_cast<const char*>(cache), length);
     std::stringstream ss;
@@ -203,7 +203,7 @@ class Int8CacheCalibrator : Algorithm {
    *
    * @return int
    */
-  int getBatchSize() const override {
+  int getBatchSize() const noexcept override {
     // HACK: TRTorch only uses explict batch sizing, INT8 Calibrator does not
     // work when reporting the batch size here and having explicity batching.
     // So we just report batch size 1 (warnings will still be printed out).
@@ -222,7 +222,7 @@ class Int8CacheCalibrator : Algorithm {
    * @param nbBindings: int - Number of bindings
    * @return false
    */
-  bool getBatch(void* bindings[], const char* names[], int nbBindings) override {
+  bool getBatch(void* bindings[], const char* names[], int nbBindings) noexcept override {
     return false;
   }
 
@@ -234,7 +234,7 @@ class Int8CacheCalibrator : Algorithm {
    * @param length
    * @return const void* - Pointer to cache data
    */
-  const void* readCalibrationCache(size_t& length) override {
+  const void* readCalibrationCache(size_t& length) noexcept override {
     std::stringstream ss;
     ss << "Reading Calibration Cache from " << cache_file_path_;
     logging::log(logging::Level::kINFO, ss.str());
@@ -258,7 +258,7 @@ class Int8CacheCalibrator : Algorithm {
    * @param cache: const void* - cache data
    * @param length: size_t - length of cache
    */
-  void writeCalibrationCache(const void* cache, size_t length) override {
+  void writeCalibrationCache(const void* cache, size_t length) noexcept override {
     std::ofstream cache_file(cache_file_path_, std::ios::binary);
     cache_file.write(reinterpret_cast<const char*>(cache), length);
     std::stringstream ss;