Prefer visqol-python 3.5.0 and ensure lattice speech support#16
Conversation
- Validated that visqol-python issue #1 is fixed (MOS consistency restored). - Updated requirements.txt to use visqol-python[accel,lattice]>=3.5.0. - Updated phase2_mos.py to prioritize visqol-python and use the deep-lattice speech model. - Updated action.yml and run_benchmark.py to make visqol-python the default backend. - Fixed Dockerfile.visqol to include lattice support and use correct PYTHONPATH. - Cleaned up visqol-python initialization logic for speech/audio modes.
|
👋 Jules, reporting for duty! I'm here to lend a hand with this pull request. When you start a review, I'll add a 👀 emoji to each comment to let you know I've read it. I'll focus on feedback directed at me and will do my best to stay out of conversations between you and other bots or reviewers to keep the noise down. I'll push a commit with your requested changes shortly after. Please note there might be a delay between these steps, but rest assured I'm on the job! For more direct control, you can switch me to Reactive Mode. When this mode is on, I will only act on comments where you specifically mention me with New to Jules? Learn more at jules.google/docs. For security, I will only act on instructions from the user who triggered this task. |
![sourcery-ai[bot]](https://avatars.githubusercontent.com/in/48477?s=60&v=4){kind=small}
- Validated visqol-python 3.5.0 speech MOS consistency against C++ binary. - Updated requirements.txt and Dockerfile.visqol to use visqol-python[accel,lattice]>=3.5.0. - Prioritized visqol-python in phase2_mos.py and run_benchmark.py auto-selection. - Added explicit support for the lattice speech model in phase2_mos.py. - Removed dead batching code from phase2_mos.py. - Updated README.md and action.yml to reflect new default backend. - Fixed Dockerfile.visqol PYTHONPATH for Ubuntu 22.04 compatibility.
…ching - Validated visqol-python 3.5.0 speech MOS consistency against C++ binary. - Updated requirements.txt to use visqol-python[accel,lattice]>=3.5.0. - Prioritized visqol-python in phase2_mos.py and run_benchmark.py auto-selection. - Implemented optimized batch processing in phase2_mos.py using measure_batch(parallel=True). - Added explicit support for the lattice speech model in phase2_mos.py. - Removed dead code and updated README.md / action.yml for new defaults. - Enhanced CI to test Docker build on PRs (build only). - Reverted Dockerfile.visqol to use visqol-py with pinned dependencies.
nschimme
force-pushed
the
prefer-visqol-python-v3.5.0-14811030729734419766
branch
from
60fd175 to
a42c51a
Compare
1 participant
This PR updates the FAAC Benchmark Suite to use
visqol-pythonas the preferred backend for perceptual quality (MOS) computation.Key changes:
visqol-python3.5.0 produces MOS scores consistent with the C++visqolbinary on the TCD-VOIP speech dataset, resolving previous concerns about inflated scores.requirements.txttovisqol-python[accel,lattice]>=3.5.0to ensure optimized execution and the presence of the deep-lattice speech mapper.phase2_mos.pyandrun_benchmark.pyto movevisqol-pythonto the top of the priority list for "auto" mode.lattice_tcditugenmeetpackhref_ls2_nl60_lr12_bs2048_learn.005_ep2400_train1_7_raw.tflitemodel for speech mode when available, ensuring parity with the C++ implementation.dockertovisqol-pythonfor faster and simpler execution.Dockerfile.visqolto include the new dependencies and fixed thePYTHONPATHfor compatibility with the Ubuntu 22.04 base image.PR created automatically by Jules for task 14811030729734419766 started by @nschimme