Support specifying decoder and its options (#2327)

Summary:
This commit adds support to specify decoder to Streamer's add stream method.
This is roughly equivalent to `ffmpeg`'s `-c:v foo` and `-c:a foo` options.

This allows to override the decoder codec and/or specify the option of
the decoder.

This change allows to specify Nvidia NVDEC codec for supported formats,
which uses dedicated hardware for decoding the video.

 ---

Note: The CL might look overwhelming, but it's essentially, add new parameters in Python, and pass them down all the way to  `AVCodecContextPtr`, which initializes the actual decoder implementation (`AVCodecContext`.)

Pull Request resolved: #2327

Differential Revision: D35626904

Pulled By: mthrok

fbshipit-source-id: 1840dd8932b4c2d62b3e3eaa0de4c50af371547b

torchaudio/csrc/ffmpeg/decoder.cpp

@@ -6,7 +6,11 @@ namespace ffmpeg {
 ////////////////////////////////////////////////////////////////////////////////
 // Decoder
 ////////////////////////////////////////////////////////////////////////////////
-Decoder::Decoder(AVCodecParameters* pParam) : pCodecContext(pParam) {}
+Decoder::Decoder(
+    AVCodecParameters* pParam,
+    const std::string& decoder_name,
+    const std::map<std::string, std::string>& decoder_option)
+    : pCodecContext(pParam, decoder_name, decoder_option) {}
 
 int Decoder::process_packet(AVPacket* pPacket) {
   return avcodec_send_packet(pCodecContext, pPacket);

torchaudio/csrc/ffmpeg/decoder.h

@@ -10,7 +10,10 @@ class Decoder {
 
  public:
   // Default constructable
-  Decoder(AVCodecParameters* pParam);
+  Decoder(
+      AVCodecParameters* pParam,
+      const std::string& decoder_name,
+      const std::map<std::string, std::string>& decoder_option);
   // Custom destructor to clean up the resources
   ~Decoder() = default;
   // Non-copyable

torchaudio/csrc/ffmpeg/ffmpeg.cpp

@@ -151,11 +151,22 @@ void AVCodecContextDeleter::operator()(AVCodecContext* p) {
 };
 
 namespace {
-AVCodecContext* get_codec_context(AVCodecParameters* pParams) {
-  const AVCodec* pCodec = avcodec_find_decoder(pParams->codec_id);
+AVCodecContext* get_codec_context(
+    enum AVCodecID codec_id,
+    const std::string& decoder_name) {
+  const AVCodec* pCodec = decoder_name.empty()
+      ? avcodec_find_decoder(codec_id)
+      : avcodec_find_decoder_by_name(decoder_name.c_str());
 
   if (!pCodec) {
-    throw std::runtime_error("Unknown codec.");
+    std::stringstream ss;
+    if (decoder_name.empty()) {
+      ss << "Unsupported codec: \"" << avcodec_get_name(codec_id) << "\", ("
+         << codec_id << ").";
+    } else {
+      ss << "Unsupported codec: \"" << decoder_name << "\".";
+    }
+    throw std::runtime_error(ss.str());
   }
 
   AVCodecContext* pCodecContext = avcodec_alloc_context3(pCodec);
@@ -167,27 +178,43 @@ AVCodecContext* get_codec_context(AVCodecParameters* pParams) {
 
 void init_codec_context(
     AVCodecContext* pCodecContext,
-    AVCodecParameters* pParams) {
-  const AVCodec* pCodec = avcodec_find_decoder(pParams->codec_id);
+    AVCodecParameters* pParams,
+    const std::string& decoder_name,
+    const std::map<std::string, std::string>& decoder_option) {
+  const AVCodec* pCodec = decoder_name.empty()
+      ? avcodec_find_decoder(pParams->codec_id)
+      : avcodec_find_decoder_by_name(decoder_name.c_str());
+
+  // No need to check if pCodec is null as it's been already checked in
+  // get_codec_context
 
   if (avcodec_parameters_to_context(pCodecContext, pParams) < 0) {
     throw std::runtime_error("Failed to set CodecContext parameter.");
   }
 
-  if (avcodec_open2(pCodecContext, pCodec, NULL) < 0) {
+  AVDictionary* opts = get_option_dict(decoder_option);
+  if (avcodec_open2(pCodecContext, pCodec, &opts) < 0) {
     throw std::runtime_error("Failed to initialize CodecContext.");
   }
+  auto unused_keys = clean_up_dict(opts);
+  if (unused_keys.size()) {
+    throw std::runtime_error(
+        "Unexpected decoder options: " + join(unused_keys));
+  }
 
   if (pParams->codec_type == AVMEDIA_TYPE_AUDIO && !pParams->channel_layout)
     pParams->channel_layout =
         av_get_default_channel_layout(pCodecContext->channels);
 }
 } // namespace
 
-AVCodecContextPtr::AVCodecContextPtr(AVCodecParameters* pParam)
+AVCodecContextPtr::AVCodecContextPtr(
+    AVCodecParameters* pParam,
+    const std::string& decoder_name,
+    const std::map<std::string, std::string>& decoder_option)
     : Wrapper<AVCodecContext, AVCodecContextDeleter>(
-          get_codec_context(pParam)) {
-  init_codec_context(ptr.get(), pParam);
+          get_codec_context(pParam->codec_id, decoder_name)) {
+  init_codec_context(ptr.get(), pParam, decoder_name, decoder_option);
 }
 ////////////////////////////////////////////////////////////////////////////////
 // AVFilterGraph

torchaudio/csrc/ffmpeg/ffmpeg.h

@@ -118,7 +118,10 @@ struct AVCodecContextDeleter {
 };
 struct AVCodecContextPtr
     : public Wrapper<AVCodecContext, AVCodecContextDeleter> {
-  AVCodecContextPtr(AVCodecParameters* pParam);
+  AVCodecContextPtr(
+      AVCodecParameters* pParam,
+      const std::string& decoder,
+      const std::map<std::string, std::string>& decoder_option);
 };
 
 ////////////////////////////////////////////////////////////////////////////////

torchaudio/csrc/ffmpeg/prototype.cpp

@@ -7,8 +7,10 @@ namespace ffmpeg {
 
 namespace {
 
+using OptionDict = c10::Dict<std::string, std::string>;
+
 std::map<std::string, std::string> convert_dict(
-    const c10::optional<c10::Dict<std::string, std::string>>& option) {
+    const c10::optional<OptionDict>& option) {
   std::map<std::string, std::string> opts;
   if (option) {
     for (auto& it : option.value()) {
@@ -22,17 +24,17 @@ struct StreamerHolder : torch::CustomClassHolder {
   Streamer s;
   StreamerHolder(
       const std::string& src,
-      c10::optional<std::string> device,
-      c10::optional<c10::Dict<std::string, std::string>> option)
+      const c10::optional<std::string>& device,
+      const c10::optional<OptionDict>& option)
       : s(src, device.value_or(""), convert_dict(option)) {}
 };
 
 using S = c10::intrusive_ptr<StreamerHolder>;
 
 S init(
     const std::string& src,
-    c10::optional<std::string> device,
-    c10::optional<c10::Dict<std::string, std::string>> option) {
+    const c10::optional<std::string>& device,
+    const c10::optional<OptionDict>& option) {
   return c10::make_intrusive<StreamerHolder>(src, device, option);
 }
 
@@ -231,7 +233,7 @@ void add_basic_audio_stream(
     const c10::optional<int64_t>& sample_rate,
     const c10::optional<c10::ScalarType>& dtype) {
   std::string filter_desc = get_afilter_desc(sample_rate, dtype);
-  s->s.add_audio_stream(i, frames_per_chunk, num_chunks, filter_desc);
+  s->s.add_audio_stream(i, frames_per_chunk, num_chunks, filter_desc, "", {});
 }
 
 void add_basic_video_stream(
@@ -244,27 +246,41 @@ void add_basic_video_stream(
     const c10::optional<int64_t>& height,
     const c10::optional<std::string>& format) {
   std::string filter_desc = get_vfilter_desc(frame_rate, width, height, format);
-  s->s.add_video_stream(i, frames_per_chunk, num_chunks, filter_desc);
+  s->s.add_video_stream(i, frames_per_chunk, num_chunks, filter_desc, "", {});
 }
 
 void add_audio_stream(
     S s,
     int64_t i,
     int64_t frames_per_chunk,
     int64_t num_chunks,
-    const c10::optional<std::string>& filter_desc) {
+    const c10::optional<std::string>& filter_desc,
+    const c10::optional<std::string>& decoder,
+    const c10::optional<OptionDict>& decoder_options) {
   s->s.add_audio_stream(
-      i, frames_per_chunk, num_chunks, filter_desc.value_or(""));
+      i,
+      frames_per_chunk,
+      num_chunks,
+      filter_desc.value_or(""),
+      decoder.value_or(""),
+      convert_dict(decoder_options));
 }
 
 void add_video_stream(
     S s,
     int64_t i,
     int64_t frames_per_chunk,
     int64_t num_chunks,
-    const c10::optional<std::string>& filter_desc) {
+    const c10::optional<std::string>& filter_desc,
+    const c10::optional<std::string>& decoder,
+    const c10::optional<OptionDict>& decoder_options) {
   s->s.add_video_stream(
-      i, frames_per_chunk, num_chunks, filter_desc.value_or(""));
+      i,
+      frames_per_chunk,
+      num_chunks,
+      filter_desc.value_or(""),
+      decoder.value_or(""),
+      convert_dict(decoder_options));
 }
 
 void remove_stream(S s, int64_t i) {
@@ -308,7 +324,7 @@ std::tuple<c10::optional<torch::Tensor>, int64_t> load(const std::string& src) {
   int i = s.find_best_audio_stream();
   auto sinfo = s.get_src_stream_info(i);
   int64_t sample_rate = static_cast<int64_t>(sinfo.sample_rate);
-  s.add_audio_stream(i, -1, -1, "");
+  s.add_audio_stream(i, -1, -1, "", "", {});
   process_all_packets(s);
   auto tensors = s.pop_chunks();
   return std::make_tuple<>(tensors[0], sample_rate);

torchaudio/csrc/ffmpeg/stream_processor.cpp

@@ -6,8 +6,11 @@ namespace ffmpeg {
 
 using KeyType = StreamProcessor::KeyType;
 
-StreamProcessor::StreamProcessor(AVCodecParameters* codecpar)
-    : decoder(codecpar) {}
+StreamProcessor::StreamProcessor(
+    AVCodecParameters* codecpar,
+    const std::string& decoder_name,
+    const std::map<std::string, std::string>& decoder_option)
+    : decoder(codecpar, decoder_name, decoder_option) {}
 
 ////////////////////////////////////////////////////////////////////////////////
 // Configurations

torchaudio/csrc/ffmpeg/stream_processor.h

@@ -25,7 +25,10 @@ class StreamProcessor {
   std::map<KeyType, Sink> sinks;
 
  public:
-  StreamProcessor(AVCodecParameters* codecpar);
+  StreamProcessor(
+      AVCodecParameters* codecpar,
+      const std::string& decoder_name,
+      const std::map<std::string, std::string>& decoder_option);
   ~StreamProcessor() = default;
   // Non-copyable
   StreamProcessor(const StreamProcessor&) = delete;

torchaudio/csrc/ffmpeg/streamer.cpp

@@ -156,39 +156,50 @@ void Streamer::add_audio_stream(
     int i,
     int frames_per_chunk,
     int num_chunks,
-    std::string filter_desc) {
+    std::string filter_desc,
+    const std::string& decoder,
+    const std::map<std::string, std::string>& decoder_option) {
   add_stream(
       i,
       AVMEDIA_TYPE_AUDIO,
       frames_per_chunk,
       num_chunks,
-      std::move(filter_desc));
+      std::move(filter_desc),
+      decoder,
+      decoder_option);
 }
 
 void Streamer::add_video_stream(
     int i,
     int frames_per_chunk,
     int num_chunks,
-    std::string filter_desc) {
+    std::string filter_desc,
+    const std::string& decoder,
+    const std::map<std::string, std::string>& decoder_option) {
   add_stream(
       i,
       AVMEDIA_TYPE_VIDEO,
       frames_per_chunk,
       num_chunks,
-      std::move(filter_desc));
+      std::move(filter_desc),
+      decoder,
+      decoder_option);
 }
 
 void Streamer::add_stream(
     int i,
     AVMediaType media_type,
     int frames_per_chunk,
     int num_chunks,
-    std::string filter_desc) {
+    std::string filter_desc,
+    const std::string& decoder,
+    const std::map<std::string, std::string>& decoder_option) {
   validate_src_stream_type(i, media_type);
   AVStream* stream = pFormatContext->streams[i];
   stream->discard = AVDISCARD_DEFAULT;
   if (!processors[i])
-    processors[i] = std::make_unique<StreamProcessor>(stream->codecpar);
+    processors[i] = std::make_unique<StreamProcessor>(
+        stream->codecpar, decoder, decoder_option);
   int key = processors[i]->add_stream(
       stream->time_base,
       stream->codecpar,

torchaudio/csrc/ffmpeg/streamer.h

@@ -66,12 +66,16 @@ class Streamer {
       int i,
       int frames_per_chunk,
       int num_chunks,
-      std::string filter_desc);
+      std::string filter_desc,
+      const std::string& decoder,
+      const std::map<std::string, std::string>& decoder_option);
   void add_video_stream(
       int i,
       int frames_per_chunk,
       int num_chunks,
-      std::string filter_desc);
+      std::string filter_desc,
+      const std::string& decoder,
+      const std::map<std::string, std::string>& decoder_option);
   void remove_stream(int i);
 
  private:
@@ -80,7 +84,9 @@ class Streamer {
       AVMediaType media_type,
       int frames_per_chunk,
       int num_chunks,
-      std::string filter_desc);
+      std::string filter_desc,
+      const std::string& decoder,
+      const std::map<std::string, std::string>& decoder_option);
 
  public:
   //////////////////////////////////////////////////////////////////////////////

torchaudio/prototype/io/streamer.py

@@ -355,6 +355,8 @@ def add_audio_stream(
         buffer_chunk_size: int = 3,
         stream_index: Optional[int] = None,
         filter_desc: Optional[str] = None,
+        decoder: Optional[str] = None,
+        decoder_options: Optional[Dict[str, str]] = None,
     ):
         """Add output audio stream
 
@@ -375,10 +377,22 @@ def add_audio_stream(
                 The list of available filters can be found at
                 https://ffmpeg.org/ffmpeg-filters.html
                 Note that complex filters are not supported.
+
+            decoder (str or None, optional): The name of the decoder to be used.
+                When provided, use the specified decoder instead of the default one.
+
+            decoder_options (dict or None, optional): Options passed to decoder.
+                Mapping from str to str.
         """
         i = self.default_audio_stream if stream_index is None else stream_index
         torch.ops.torchaudio.ffmpeg_streamer_add_audio_stream(
-            self._s, i, frames_per_chunk, buffer_chunk_size, filter_desc
+            self._s,
+            i,
+            frames_per_chunk,
+            buffer_chunk_size,
+            filter_desc,
+            decoder,
+            decoder_options,
         )
 
     def add_video_stream(
@@ -387,6 +401,8 @@ def add_video_stream(
         buffer_chunk_size: int = 3,
         stream_index: Optional[int] = None,
         filter_desc: Optional[str] = None,
+        decoder: Optional[str] = None,
+        decoder_options: Optional[Dict[str, str]] = None,
     ):
         """Add output video stream
 
@@ -407,10 +423,22 @@ def add_video_stream(
                 The list of available filters can be found at
                 https://ffmpeg.org/ffmpeg-filters.html
                 Note that complex filters are not supported.
+
+            decoder (str or None, optional): The name of the decoder to be used.
+                When provided, use the specified decoder instead of the default one.
+
+            decoder_options (dict or None, optional): Options passed to decoder.
+                Mapping from str to str.
         """
         i = self.default_video_stream if stream_index is None else stream_index
         torch.ops.torchaudio.ffmpeg_streamer_add_video_stream(
-            self._s, i, frames_per_chunk, buffer_chunk_size, filter_desc
+            self._s,
+            i,
+            frames_per_chunk,
+            buffer_chunk_size,
+            filter_desc,
+            decoder,
+            decoder_options,
         )
 
     def remove_stream(self, i: int):