fixes to pass metadata tests

Author: Dan-Flores

output.mp3

@@ -479,10 +479,10 @@ std::string get_json_metadata(at::Tensor& decoder) {
     auto streamMetadata =
         videoMetadata.allStreamMetadata[*maybeBestVideoStreamIndex];
     if (streamMetadata.numFramesFromScan.has_value()) {
-      metadataMap["numFrames"] =
+      metadataMap["numFramesFromHeader"] =
           std::to_string(*streamMetadata.numFramesFromScan);
     } else if (streamMetadata.numFramesFromHeader.has_value()) {
-      metadataMap["numFrames"] =
+      metadataMap["numFramesFromHeader"] =
           std::to_string(*streamMetadata.numFramesFromHeader);
     }
     if (streamMetadata.minPtsSecondsFromScan.has_value()) {

src/torchcodec/_core/custom_ops.cpp

@@ -213,7 +213,7 @@ def _get_clips_for_index_based_sampling(
         sample_end_index = (
             min(
                 index_based_sampler_args.sample_end_index + 1,
-                metadata_json["numFrames"],
+                metadata_json["numFramesFromHeader"],
             )
             - index_based_sampler_args.video_frame_dilation
             * index_based_sampler_args.frames_per_clip
@@ -263,13 +263,13 @@ def _get_start_seconds(
         Returns:
             (`List[float]`): List of the sampled clip start position in seconds
         """
-        video_duration_in_seconds = metadata_json["durationSeconds"]
+        video_duration_in_seconds = metadata_json["durationSecondsFromHeader"]
 
         clip_duration_in_seconds = (
             time_based_sampler_args.frames_per_clip
             * time_based_sampler_args.video_frame_dilation
             + 1
-        ) / metadata_json["averageFps"]
+        ) / metadata_json["averageFpsFromHeader"]
 
         minPtsSecondsFromScan = (
             metadata_json["minPtsSecondsFromScan"]

src/torchcodec/_samplers/video_clip_sampler.py

@@ -215,7 +215,7 @@ def test_pts_apis_against_index_ref(self, device):
 
         metadata = get_json_metadata(decoder)
         metadata_dict = json.loads(metadata)
-        num_frames = metadata_dict["numFrames"]
+        num_frames = metadata_dict["numFramesFromHeader"]
         assert num_frames == 390
 
         _, all_pts_seconds_ref, _ = zip(
@@ -395,9 +395,11 @@ def test_video_get_json_metadata(self):
         metadata_dict = json.loads(metadata)
 
         # We should be able to see all of this metadata without adding a video stream
-        assert metadata_dict["durationSeconds"] == pytest.approx(13.013, abs=0.001)
-        assert metadata_dict["numFrames"] == 390
-        assert metadata_dict["averageFps"] == pytest.approx(29.97, abs=0.001)
+        assert metadata_dict["durationSecondsFromHeader"] == pytest.approx(
+            13.013, abs=0.001
+        )
+        assert metadata_dict["numFramesFromHeader"] == 390
+        assert metadata_dict["averageFpsFromHeader"] == pytest.approx(29.97, abs=0.001)
         assert metadata_dict["codec"] == "h264"
         ffmpeg_dict = get_ffmpeg_library_versions()
         if ffmpeg_dict["libavformat"][0] >= 60: