[VLM] Replace decord with torchcodec for video decoding (#20055)

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
Co-authored-by: BakerBunker <17872844+BakerBunker@users.noreply.github.com>

python/sglang/test/vlm_utils.py

@@ -378,20 +378,16 @@ class ImageOpenAITestMixin(TestOpenAIMLLMServerBase):
         # the memory consumed by the Vision Attention varies a lot, e.g. blocked qkv vs full-sequence sdpa
         # the size of the video embeds differs from the `modality` argument when preprocessed
 
-        # We import decord here to avoid a strange Segmentation fault (core dumped) issue.
-        # The following import order will cause Segmentation fault.
-        # import decord
-        # from transformers import AutoTokenizer
-        from decord import VideoReader, cpu
+        from sglang.srt.utils.video_decoder import VideoDecoderWrapper
 
         max_frames_num = 10
-        vr = VideoReader(video_path, ctx=cpu(0))
-        total_frame_num = len(vr)
+        decoder = VideoDecoderWrapper(video_path)
+        total_frame_num = len(decoder)
         uniform_sampled_frames = np.linspace(
             0, total_frame_num - 1, max_frames_num, dtype=int
         )
         frame_idx = uniform_sampled_frames.tolist()
-        frames = vr.get_batch(frame_idx).asnumpy()
+        frames = decoder.get_frames_at(frame_idx)
 
         base64_frames = []
         for frame in frames: