# Latest-dependency CosyVoice stack for the Ubuntu 26.04 Quadro M4000 CUDA lane.
#
# Keep the stack loose so pip can take current compatible releases, but cap
# transformers below 4.53.0. Matrix verification in the Arc B50 and M4000
# workspaces showed:
# - transformers 4.52.4: generated the requested text correctly
# - transformers 4.53.0 and newer tested through 4.57.6: GPU run completed,
#   but generated semantically broken speech unrelated to the input text
huggingface_hub>=0.30.0
conformer>=0.3.2
diffusers>=0.29.0
gradio>=5.4.0
gdown>=5.1.0
hydra-core>=1.3.2
HyperPyYAML>=1.2.3
inflect>=7.3.1
librosa>=0.10.2
lightning>=2.2.4
matplotlib>=3.7.5
modelscope>=1.20.0
networkx>=3.1
numpy>=1.26.4
omegaconf>=2.3.0
onnx>=1.16.0
onnxruntime>=1.18.0
openai-whisper>=20250625
pyarrow>=18.1.0
pyworld>=0.3.4
rich>=13.7.1
soundfile>=0.12.1
torchcodec>=0.13,<0.14
transformers>=4.51.3,<4.53
x-transformers>=2.11.24
wetext>=0.0.4
wget>=3.2