Merge branch 'main' of github.com:pytorch/torchcodec into approx

Author: scotts

src/torchcodec/decoders/_core/CPUOnlyDevice.cpp

@@ -35,7 +35,7 @@ void releaseContextOnCuda(
   throwUnsupportedDeviceError(device);
 }
 
-std::optional<AVCodecPtr> findCudaCodec(
+std::optional<const AVCodec*> findCudaCodec(
     const torch::Device& device,
     [[maybe_unused]] const AVCodecID& codecId) {
   throwUnsupportedDeviceError(device);

src/torchcodec/decoders/_core/DeviceInterface.h

@@ -40,7 +40,7 @@ void releaseContextOnCuda(
     const torch::Device& device,
     AVCodecContext* codecContext);
 
-std::optional<AVCodecPtr> findCudaCodec(
+std::optional<const AVCodec*> findCudaCodec(
     const torch::Device& device,
     const AVCodecID& codecId);
 

src/torchcodec/decoders/_core/FFMPEGCommon.cpp

@@ -10,6 +10,15 @@
 
 namespace facebook::torchcodec {
 
+AVCodecOnlyUseForCallingAVFindBestStream
+makeAVCodecOnlyUseForCallingAVFindBestStream(const AVCodec* codec) {
+#if LIBAVCODEC_VERSION_INT < AV_VERSION_INT(59, 18, 100)
+  return const_cast<AVCodec*>(codec);
+#else
+  return codec;
+#endif
+}
+
 std::string getFFMPEGErrorStringFromErrorCode(int errorCode) {
   char errorBuffer[AV_ERROR_MAX_STRING_SIZE] = {0};
   av_strerror(errorCode, errorBuffer, AV_ERROR_MAX_STRING_SIZE);

src/torchcodec/decoders/_core/FFMPEGCommon.h

@@ -75,12 +75,17 @@ using UniqueSwsContext =
 // which was released in FFMPEG version=5.0.3
 // with libavcodec's version=59.18.100
 // (https://www.ffmpeg.org/olddownload.html).
+// Note that the alias is so-named so that it is only used when interacting with
+// av_find_best_stream(). It is not needed elsewhere.
 #if LIBAVCODEC_VERSION_INT < AV_VERSION_INT(59, 18, 100)
-using AVCodecPtr = AVCodec*;
+using AVCodecOnlyUseForCallingAVFindBestStream = AVCodec*;
 #else
-using AVCodecPtr = const AVCodec*;
+using AVCodecOnlyUseForCallingAVFindBestStream = const AVCodec*;
 #endif
 
+AVCodecOnlyUseForCallingAVFindBestStream
+makeAVCodecOnlyUseForCallingAVFindBestStream(const AVCodec* codec);
+
 // Success code from FFMPEG is just a 0. We define it to make the code more
 // readable.
 const int AVSUCCESS = 0;

src/torchcodec/decoders/_core/VideoDecoder.cpp

@@ -436,7 +436,7 @@ void VideoDecoder::createFilterGraph(
 }
 
 int VideoDecoder::getBestStreamIndex(AVMediaType mediaType) {
-  AVCodecPtr codec = nullptr;
+  AVCodecOnlyUseForCallingAVFindBestStream codec = nullptr;
   int streamNumber =
       av_find_best_stream(formatContext_.get(), mediaType, -1, -1, &codec, 0);
   return streamNumber;
@@ -452,7 +452,7 @@ void VideoDecoder::addVideoStreamDecoder(
   }
   TORCH_CHECK(formatContext_.get() != nullptr);
 
-  AVCodecPtr codec = nullptr;
+  AVCodecOnlyUseForCallingAVFindBestStream codec = nullptr;
   int streamNumber = av_find_best_stream(
       formatContext_.get(),
       AVMEDIA_TYPE_VIDEO,
@@ -465,14 +465,6 @@ void VideoDecoder::addVideoStreamDecoder(
   }
   TORCH_CHECK(codec != nullptr);
 
-  StreamMetadata& streamMetadata = containerMetadata_.streams[streamNumber];
-  if (seekMode_ == SeekMode::approximate &&
-      !streamMetadata.averageFps.has_value()) {
-    throw std::runtime_error(
-        "Seek mode is approximate, but stream " + std::to_string(streamNumber) +
-        " does not have an average fps in its metadata.");
-  }
-
   StreamInfo& streamInfo = streams_[streamNumber];
   streamInfo.streamIndex = streamNumber;
   streamInfo.timeBase = formatContext_->streams[streamNumber]->time_base;
@@ -485,8 +477,17 @@ void VideoDecoder::addVideoStreamDecoder(
   }
 
   if (options.device.type() == torch::kCUDA) {
-    codec = findCudaCodec(options.device, streamInfo.stream->codecpar->codec_id)
-                .value_or(codec);
+    codec = makeAVCodecOnlyUseForCallingAVFindBestStream(
+        findCudaCodec(options.device, streamInfo.stream->codecpar->codec_id)
+            .value_or(codec));
+  }
+
+  StreamMetadata& streamMetadata = containerMetadata_.streams[streamNumber];
+  if (seekMode_ == SeekMode::approximate &&
+      !streamMetadata.averageFps.has_value()) {
+    throw std::runtime_error(
+        "Seek mode is approximate, but stream " + std::to_string(streamNumber) +
+        " does not have an average fps in its metadata.");
   }
 
   AVCodecContext* codecContext = avcodec_alloc_context3(codec);

test/decoders/test_video_decoder.py

@@ -428,6 +428,9 @@ def test_get_frames_at_fails(self, device, seek_mode):
 
     @pytest.mark.parametrize("device", cpu_and_cuda())
     def test_get_frame_at_av1(self, device):
+        if in_fbcode() and device == "cuda":
+            return
+
         decoder = VideoDecoder(AV1_VIDEO.path, device=device)
         ref_frame10 = AV1_VIDEO.get_frame_data_by_index(10)
         ref_frame_info10 = AV1_VIDEO.get_frame_info(10)