PyAnnote environment installs torchcodec 0.17, which fails to load next to torch 2.8 (AudioDecoder is not defined)
Nobody has claimed this yet.
Assessment
- Difficulty
- 2/5
- Estimated time
- 1-3 hours
- Newbie friendliness
- 78/100
- Issue type
- Bug
- Clarity
- Clearly specified
- Activity status
- Active
- Domain
- audio-video-rtc, machine-learning
Research direction
Start with internal/transcription/adapters/py/pyannote/pyproject.toml, where the issue says torchcodec is unpinned; compare its dependency setup with the WhisperX pin in internal/transcription/adapters/whisperx_adapter.go and check the related reports #448 and #425. Verify the PyAnnote environment resolves a compatible torchcodec alongside torch 2.8, then confirm fresh Parakeet and Canary diarization jobs work without the AudioDecoder error.
Written by the indexing model from the issue text.
Description
Summary
The PyAnnote diarization environment resolves torchcodec 0.17.0 next to torch 2.8.0+cu128. torchcodec 0.17 is built for a newer CUDA, fails to load, and diarization stops with name 'AudioDecoder' is not defined.
Environment
- Scriberr built from
mainat a353078 withDockerfile.cuda.12.9 - Docker 29.7.2 on x86_64 (Linux), NVIDIA RTX PRO 6000 Blackwell, driver 595.91.07
- Fresh named volumes for
/app/dataand/app/whisperx-env
Actual behavior
Error during diarization: name 'AudioDecoder' is not defined
Importing torchcodec directly in that environment shows the cause.
OSError: libnvrtc.so.13: cannot open shared object file: No such file or directory
bdb8838 pins torchcodec~=0.7.0 for the WhisperX environment only (whisperx_adapter.go#L355-L357). The PyAnnote environment lists torchcodec without a version (py/pyannote/pyproject.toml#L20-L21), so it gets the newest release.
Expected behavior
pyannote diarization works for Parakeet and Canary jobs on a freshly built environment.
Workaround
Pinning torchcodec in the PyAnnote environment and relocking fixes it.
override-dependencies = [
"torchcodec~=0.7.0 ; sys_platform != 'linux' or platform_machine != 'aarch64'"
]
#448 and #425 show the same torchcodec loading error in the WhisperX environment on other platforms.
- Dominant language
- Go
- Stars
- 3.1k
- Forks
- 272
- PR merge metrics
- No merged PRs in 30d
Getting set up
- Ships a Dockerfile or Docker Compose file
- No pull request template
- No contributing guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
More from rishikanthc/Scriberr
-
Difficulty 2/5 1-3 hours Newbie friendliness 73/100
rishikanthc/Scriberr#507 ·
-
Difficulty 2/5 1-3 hours Newbie friendliness 80/100
rishikanthc/Scriberr#506 ·
-
Difficulty 2/5 1-3 hours Newbie friendliness 73/100
rishikanthc/Scriberr#497 ·
-
API start endpoint requires hf_token for Parakeet and Canary diarization even when HF_TOKEN is setOpen
Difficulty 2/5 1-3 hours Newbie friendliness 77/100
rishikanthc/Scriberr#495 ·
-
Difficulty 2/5 1-3 hours Newbie friendliness 68/100
rishikanthc/Scriberr#456 · 1 comment ·
All issues in rishikanthc/Scriberr
Similar issues
-
Difficulty 2/5 1-3 hours Newbie friendliness 68/100
dragonflydb/dragonfly-operator#619 ·
Maintainers usually reply within 2 days
-
Difficulty 2/5 1-3 hours Newbie friendliness 67/100
cloudfoundry/bosh-bootloader#701 · 1 comment ·
-
Difficulty 2/5 1-3 hours Newbie friendliness 76/100
-
baremetal: private network check rejects offer-specific options (Private Network 10Gbps / 25Gbps)Possibly taken @lbarrosby claimed this today. Open
Difficulty 2/5 1-3 hours Newbie friendliness 70/100
scaleway/terraform-provider-scaleway#4587 ·
Maintainers usually reply within 1 day
-
blocked-needs-validation bug
Difficulty 1/5 Under an hour Newbie friendliness 78/100
projectcapsule/capsule#2280 ·
Maintainers usually reply within 1 day