decord reads video but cannot use the GPU: distinguish the installed build
Check the Python binding, media input and native decoding capability separately. CUDA availability in Torch does not establish GPU decoding support in Decord.
Scope and symptoms
Check the Python binding, media input and native decoding capability separately. CUDA availability in Torch does not establish GPU decoding support in Decord.
Search fragments; wording and context vary:
No module named 'decord'
CUDA not enabled
Source-supported context
The Decord README describes a CPU package distribution and a separate source-build route for GPU support. Apply that information to the actual release rather than copying old platform-specific build instructions indiscriminately. V04-DECORD
Distinguish these cases
1. The binding is absent from the intended interpreter.
2. CPU reading works, while the node requests a GPU context unavailable in that build.
3. Decoding works generally but a particular file is damaged, inaccessible or unsupported.
Suggested diagnostic sequence
This sequence is editorial guidance, not a diagnosis of your machine.
1. Record the node, Decord version, platform and installation source. Locate the failure at import, reader creation or device selection.
2. Test CPU decoding and one frame from a short trusted video first. A single-file failure calls for media and access checks before changing CUDA components.
3. Use a documented CPU path where the node permits one. If GPU decoding is required, follow the relevant release's native-build and device requirements in an isolated environment.
4. After the device path works, check the node's tensor bridge, frame count and time-index assumptions. Repeat the same small clip before scaling to a full task.
Completion checks
The target file is read on the intended path and the frame count and content are correct. Torch reporting a GPU, or a faster run, does not by itself verify the Decord build.
Limits and cautions
Do not reuse obsolete operating-system repositories or compiler commands from historical instructions without checking applicability. GPU decoding and GPU model inference are independent settings.
Source review: 2026-09-22. No installation, GPU inference or user-environment repair was executed.
Original sources
- Decord upstream README · 2026-09-22
Related guides
Was this page helpful?
Anonymous. Only a yes/no counter is stored; no account, IP address or device details.
Sources & references
Source-based editorial draft; no runtime verification. Version observations are scoped, and unknown wrapper identity is explicit.
01Decord upstream READMESource checked: 2026-09-22