Merge branch 'main' into main

docs/source/libtorio.stream_reader.rst

@@ -24,7 +24,7 @@ StreamingMediaDecoder
 
 .. doxygenclass:: torio::io::StreamingMediaDecoder
 
-.. doxygenfunction:: torio::io::StreamingMediaDecoder::StreamingMediaDecoder(const std::string &src, const c10::optional<std::string> &format = {}, const c10::optional<OptionDict> &option = {})
+.. doxygenfunction:: torio::io::StreamingMediaDecoder::StreamingMediaDecoder(const std::string &src, const std::optional<std::string> &format = {}, const c10::optional<OptionDict> &option = {})
 
 StreamingMediaDecoderCustomIO
 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

docs/source/libtorio.stream_writer.rst

@@ -24,7 +24,7 @@ StreamingMediaEncoder
 
 .. doxygenclass:: torio::io::StreamingMediaEncoder
 
-.. doxygenfunction:: torio::io::StreamingMediaEncoder::StreamingMediaEncoder(const std::string &dst, const c10::optional<std::string> &format = {})
+.. doxygenfunction:: torio::io::StreamingMediaEncoder::StreamingMediaEncoder(const std::string &dst, const std::optional<std::string> &format = {})
 
 StreamingMediaEncoderCustomIO
 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

src/libtorchaudio/rnnt/autograd.cpp

@@ -42,7 +42,7 @@ class RNNTLossFunction : public torch::autograd::Function<RNNTLossFunction> {
   }
 };
 
-std::tuple<torch::Tensor, c10::optional<torch::Tensor>> rnnt_loss_autograd(
+std::tuple<torch::Tensor, std::optional<torch::Tensor>> rnnt_loss_autograd(
     torch::Tensor& logits,
     const torch::Tensor& targets,
     const torch::Tensor& logit_lengths,

src/libtorchaudio/rnnt/compute.cpp

@@ -1,7 +1,7 @@
 #include <libtorchaudio/rnnt/compute.h>
 #include <torch/script.h>
 
-std::tuple<torch::Tensor, c10::optional<torch::Tensor>> rnnt_loss(
+std::tuple<torch::Tensor, std::optional<torch::Tensor>> rnnt_loss(
     torch::Tensor& logits,
     const torch::Tensor& targets,
     const torch::Tensor& logit_lengths,

src/libtorchaudio/rnnt/compute.h

@@ -2,7 +2,7 @@
 
 #include <torch/script.h>
 
-std::tuple<torch::Tensor, c10::optional<torch::Tensor>> rnnt_loss(
+std::tuple<torch::Tensor, std::optional<torch::Tensor>> rnnt_loss(
     torch::Tensor& logits,
     const torch::Tensor& targets,
     const torch::Tensor& logit_lengths,

src/libtorchaudio/rnnt/cpu/compute.cpp

@@ -6,7 +6,7 @@ namespace rnnt {
 namespace cpu {
 
 // Entry point into RNNT Loss
-std::tuple<torch::Tensor, c10::optional<torch::Tensor>> compute(
+std::tuple<torch::Tensor, std::optional<torch::Tensor>> compute(
     torch::Tensor& logits,
     const torch::Tensor& targets,
     const torch::Tensor& logit_lengths,
@@ -89,7 +89,7 @@ std::tuple<torch::Tensor, c10::optional<torch::Tensor>> compute(
   torch::Tensor costs = torch::empty(
       options.batchSize_ * options.nHypos_,
       torch::TensorOptions().device(logits.device()).dtype(logits.dtype()));
-  c10::optional<torch::Tensor> gradients = torch::zeros_like(logits);
+  std::optional<torch::Tensor> gradients = torch::zeros_like(logits);
 
   torch::Tensor int_workspace = torch::empty(
       IntWorkspace::ComputeSizeFromOptions(options),

src/libtorchaudio/rnnt/gpu/compute.cu

@@ -7,7 +7,7 @@ namespace rnnt {
 namespace gpu {
 
 // Entry point into RNNT Loss
-std::tuple<torch::Tensor, c10::optional<torch::Tensor>> compute(
+std::tuple<torch::Tensor, std::optional<torch::Tensor>> compute(
     torch::Tensor& logits,
     const torch::Tensor& targets,
     const torch::Tensor& logit_lengths,
@@ -92,7 +92,7 @@ std::tuple<torch::Tensor, c10::optional<torch::Tensor>> compute(
   torch::Tensor costs = torch::empty(
       options.batchSize_ * options.nHypos_,
       torch::TensorOptions().device(logits.device()).dtype(logits.dtype()));
-  c10::optional<torch::Tensor> gradients = torch::zeros_like(logits);
+  std::optional<torch::Tensor> gradients = torch::zeros_like(logits);
 
   torch::Tensor int_workspace = torch::empty(
       IntWorkspace::ComputeSizeFromOptions(options),

src/libtorchaudio/sox/effects.cpp

@@ -86,9 +86,9 @@ auto apply_effects_tensor(
 auto apply_effects_file(
     const std::string& path,
     const std::vector<std::vector<std::string>>& effects,
-    c10::optional<bool> normalize,
-    c10::optional<bool> channels_first,
-    const c10::optional<std::string>& format)
+    std::optional<bool> normalize,
+    std::optional<bool> channels_first,
+    const std::optional<std::string>& format)
     -> std::tuple<torch::Tensor, int64_t> {
   // Open input file
   SoxFormat sf(sox_open_read(

src/libtorchaudio/sox/effects.h

@@ -19,9 +19,9 @@ auto apply_effects_tensor(
 auto apply_effects_file(
     const std::string& path,
     const std::vector<std::vector<std::string>>& effects,
-    c10::optional<bool> normalize,
-    c10::optional<bool> channels_first,
-    const c10::optional<std::string>& format)
+    std::optional<bool> normalize,
+    std::optional<bool> channels_first,
+    const std::optional<std::string>& format)
     -> std::tuple<torch::Tensor, int64_t>;
 
 } // namespace torchaudio::sox

src/libtorchaudio/sox/io.cpp

@@ -10,7 +10,7 @@ namespace torchaudio::sox {
 
 std::tuple<int64_t, int64_t, int64_t, int64_t, std::string> get_info_file(
     const std::string& path,
-    const c10::optional<std::string>& format) {
+    const std::optional<std::string>& format) {
   SoxFormat sf(sox_open_read(
       path.c_str(),
       /*signal=*/nullptr,
@@ -28,8 +28,8 @@ std::tuple<int64_t, int64_t, int64_t, int64_t, std::string> get_info_file(
 }
 
 std::vector<std::vector<std::string>> get_effects(
-    const c10::optional<int64_t>& frame_offset,
-    const c10::optional<int64_t>& num_frames) {
+    const std::optional<int64_t>& frame_offset,
+    const std::optional<int64_t>& num_frames) {
   const auto offset = frame_offset.value_or(0);
   TORCH_CHECK(
       offset >= 0,
@@ -57,11 +57,11 @@ std::vector<std::vector<std::string>> get_effects(
 
 std::tuple<torch::Tensor, int64_t> load_audio_file(
     const std::string& path,
-    const c10::optional<int64_t>& frame_offset,
-    const c10::optional<int64_t>& num_frames,
-    c10::optional<bool> normalize,
-    c10::optional<bool> channels_first,
-    const c10::optional<std::string>& format) {
+    const std::optional<int64_t>& frame_offset,
+    const std::optional<int64_t>& num_frames,
+    std::optional<bool> normalize,
+    std::optional<bool> channels_first,
+    const std::optional<std::string>& format) {
   auto effects = get_effects(frame_offset, num_frames);
   return apply_effects_file(path, effects, normalize, channels_first, format);
 }
@@ -71,10 +71,10 @@ void save_audio_file(
     torch::Tensor tensor,
     int64_t sample_rate,
     bool channels_first,
-    c10::optional<double> compression,
-    c10::optional<std::string> format,
-    c10::optional<std::string> encoding,
-    c10::optional<int64_t> bits_per_sample) {
+    std::optional<double> compression,
+    std::optional<std::string> format,
+    std::optional<std::string> encoding,
+    std::optional<int64_t> bits_per_sample) {
   validate_input_tensor(tensor);
 
   const auto filetype = [&]() {

src/libtorchaudio/sox/io.h

@@ -7,31 +7,31 @@
 namespace torchaudio::sox {
 
 auto get_effects(
-    const c10::optional<int64_t>& frame_offset,
-    const c10::optional<int64_t>& num_frames)
+    const std::optional<int64_t>& frame_offset,
+    const std::optional<int64_t>& num_frames)
     -> std::vector<std::vector<std::string>>;
 
 std::tuple<int64_t, int64_t, int64_t, int64_t, std::string> get_info_file(
     const std::string& path,
-    const c10::optional<std::string>& format);
+    const std::optional<std::string>& format);
 
 std::tuple<torch::Tensor, int64_t> load_audio_file(
     const std::string& path,
-    const c10::optional<int64_t>& frame_offset,
-    const c10::optional<int64_t>& num_frames,
-    c10::optional<bool> normalize,
-    c10::optional<bool> channels_first,
-    const c10::optional<std::string>& format);
+    const std::optional<int64_t>& frame_offset,
+    const std::optional<int64_t>& num_frames,
+    std::optional<bool> normalize,
+    std::optional<bool> channels_first,
+    const std::optional<std::string>& format);
 
 void save_audio_file(
     const std::string& path,
     torch::Tensor tensor,
     int64_t sample_rate,
     bool channels_first,
-    c10::optional<double> compression,
-    c10::optional<std::string> format,
-    c10::optional<std::string> encoding,
-    c10::optional<int64_t> bits_per_sample);
+    std::optional<double> compression,
+    std::optional<std::string> format,
+    std::optional<std::string> encoding,
+    std::optional<int64_t> bits_per_sample);
 
 } // namespace torchaudio::sox
 

src/libtorchaudio/sox/types.cpp

@@ -67,7 +67,7 @@ std::string to_string(Encoding v) {
   }
 }
 
-Encoding get_encoding_from_option(const c10::optional<std::string>& encoding) {
+Encoding get_encoding_from_option(const std::optional<std::string>& encoding) {
   if (!encoding.has_value())
     return Encoding::NOT_PROVIDED;
   std::string v = encoding.value();
@@ -84,7 +84,7 @@ Encoding get_encoding_from_option(const c10::optional<std::string>& encoding) {
   TORCH_CHECK(false, "Internal Error: unexpected encoding value: ", v);
 }
 
-BitDepth get_bit_depth_from_option(const c10::optional<int64_t>& bit_depth) {
+BitDepth get_bit_depth_from_option(const std::optional<int64_t>& bit_depth) {
   if (!bit_depth.has_value())
     return BitDepth::NOT_PROVIDED;
   int64_t v = bit_depth.value();

src/libtorchaudio/sox/types.h

@@ -38,7 +38,7 @@ enum class Encoding {
 };
 
 std::string to_string(Encoding v);
-Encoding get_encoding_from_option(const c10::optional<std::string>& encoding);
+Encoding get_encoding_from_option(const std::optional<std::string>& encoding);
 
 enum class BitDepth : unsigned {
   NOT_PROVIDED = 0,
@@ -49,7 +49,7 @@ enum class BitDepth : unsigned {
   B64 = 64,
 };
 
-BitDepth get_bit_depth_from_option(const c10::optional<int64_t>& bit_depth);
+BitDepth get_bit_depth_from_option(const std::optional<int64_t>& bit_depth);
 
 std::string get_encoding(sox_encoding_t encoding);
 

src/libtorchaudio/sox/utils.cpp

@@ -289,8 +289,8 @@ std::tuple<sox_encoding_t, unsigned> get_save_encoding_for_wav(
 std::tuple<sox_encoding_t, unsigned> get_save_encoding(
     const std::string& format,
     const caffe2::TypeMeta& dtype,
-    const c10::optional<std::string>& encoding,
-    const c10::optional<int64_t>& bits_per_sample) {
+    const std::optional<std::string>& encoding,
+    const std::optional<int64_t>& bits_per_sample) {
   const Format fmt = get_format_from_string(format);
   const Encoding enc = get_encoding_from_option(encoding);
   const BitDepth bps = get_bit_depth_from_option(bits_per_sample);
@@ -492,9 +492,9 @@ sox_encodinginfo_t get_tensor_encodinginfo(caffe2::TypeMeta dtype) {
 sox_encodinginfo_t get_encodinginfo_for_save(
     const std::string& format,
     const caffe2::TypeMeta& dtype,
-    const c10::optional<double>& compression,
-    const c10::optional<std::string>& encoding,
-    const c10::optional<int64_t>& bits_per_sample) {
+    const std::optional<double>& compression,
+    const std::optional<std::string>& encoding,
+    const std::optional<int64_t>& bits_per_sample) {
   auto enc = get_save_encoding(format, dtype, encoding, bits_per_sample);
   return sox_encodinginfo_t{
       /*encoding=*/std::get<0>(enc),

src/libtorchaudio/sox/utils.h

@@ -104,9 +104,9 @@ sox_encodinginfo_t get_tensor_encodinginfo(const caffe2::TypeMeta dtype);
 sox_encodinginfo_t get_encodinginfo_for_save(
     const std::string& format,
     const caffe2::TypeMeta& dtype,
-    const c10::optional<double>& compression,
-    const c10::optional<std::string>& encoding,
-    const c10::optional<int64_t>& bits_per_sample);
+    const std::optional<double>& compression,
+    const std::optional<std::string>& encoding,
+    const std::optional<int64_t>& bits_per_sample);
 
 } // namespace torchaudio::sox
 #endif

src/libtorchaudio/utils.cpp

@@ -23,7 +23,7 @@ bool is_align_available() {
 #endif
 }
 
-c10::optional<int64_t> cuda_version() {
+std::optional<int64_t> cuda_version() {
 #ifdef USE_CUDA
   return CUDA_VERSION;
 #else

src/libtorchaudio/utils.h

@@ -4,5 +4,5 @@
 namespace torchaudio {
 bool is_rir_available();
 bool is_align_available();
-c10::optional<int64_t> cuda_version();
+std::optional<int64_t> cuda_version();
 } // namespace torchaudio

src/libtorio/ffmpeg/ffmpeg.cpp

@@ -10,7 +10,7 @@ namespace torio::io {
 ////////////////////////////////////////////////////////////////////////////////
 // AVDictionary
 ////////////////////////////////////////////////////////////////////////////////
-AVDictionary* get_option_dict(const c10::optional<OptionDict>& option) {
+AVDictionary* get_option_dict(const std::optional<OptionDict>& option) {
   AVDictionary* opt = nullptr;
   if (option) {
     for (auto const& [key, value] : option.value()) {

src/libtorio/ffmpeg/ffmpeg.h

@@ -77,7 +77,7 @@ class Wrapper {
 // IIRC-semantic. Instead we provide helper functions.
 
 // Convert standard dict to FFmpeg native type
-AVDictionary* get_option_dict(const c10::optional<OptionDict>& option);
+AVDictionary* get_option_dict(const std::optional<OptionDict>& option);
 
 // Clean up the dict after use. If there is an unsed key, throw runtime error
 void clean_up_dict(AVDictionary* p);