13 lines
192 B
Plaintext
13 lines
192 B
Plaintext
torch
|
|
numpy
|
|
transformers==4.52.3
|
|
accelerate
|
|
qwen-omni-utils[decord]
|
|
librosa
|
|
Pillow
|
|
tqdm
|
|
scikit-learn
|
|
|
|
# Optional on compatible NVIDIA GPUs: flash-attn
|
|
# Optional NVDEC backend: PyNvVideoCodec
|