NeptuneHub/AudioMuse-AI v2.0.0-model

Model v2.0.0 Release: ONNX model conversion

on GitHub

This is the second release of model, it convert the .pb model to .onnx in order to have a more standardized result across different CPU and Library version.

The original .pb model come from:

• https://essentia.upf.edu/models.html

Was then converted by this code:

#!/bin/bash
set -e

# --- The download section has been commented out as you already have the model files. ---
# echo "--- Downloading all TensorFlow models (.pb) ---"
# wget -q \
#     https://github.com/NeptuneHub/AudioMuse-AI/releases/download/v1.0.0-model/danceability-msd-musicnn-1.pb \
#     https://github.com/NeptuneHub/AudioMuse-AI/releases/download/v1.0.0-model/mood_aggressive-msd-musicnn-1.pb \
#     https://github.com/NeptuneHub/AudioMuse-AI/releases/download/v1.0.0-model/mood_happy-msd-musicnn-1.pb \
#     https://github.com/NeptuneHub/AudioMuse-AI/releases/download/v1.0.0-model/mood_party-msd-musicnn-1.pb \
#     https://github.com/NeptuneHub/AudioMuse-AI/releases/download/v1.0.0-model/mood_relaxed-msd-musicnn-1.pb \
#     https://github.com/NeptuneHub/AudioMuse-AI/releases/download/v1.0.0-model/mood_sad-msd-musicnn-1.pb \
#     https://github.com/NeptuneHub/AudioMuse-AI/releases/download/v1.0.0-model/msd-msd-musicnn-1.pb \
#     https://github.com/NeptuneHub/AudioMuse-AI/releases/download/v1.0.0-model/msd-musicnn-1.pb

echo "--- Starting model conversions to ONNX ---"

echo "Converting embedding model (msd-musicnn-1.pb)..."
python -m tf2onnx.convert \
  --input ./msd-musicnn-1.pb \
  --inputs 'model/Placeholder:0' \
  --outputs 'model/dense/BiasAdd:0' \
  --output msd-musicnn-1.onnx

echo "Converting prediction model (msd-msd-musicnn-1.pb)..."
python -m tf2onnx.convert \
  --input ./msd-msd-musicnn-1.pb \
  --inputs 'serving_default_model_Placeholder:0' \
  --outputs 'PartitionedCall:0' \
  --output msd-msd-musicnn-1.onnx

echo "Converting danceability model (danceability-msd-musicnn-1.pb)..."
python -m tf2onnx.convert \
  --input ./danceability-msd-musicnn-1.pb \
  --inputs 'model/Placeholder:0' \
  --outputs 'model/Softmax:0' \
  --output danceability-msd-musicnn-1.onnx

echo "Converting aggressive mood model (mood_aggressive-msd-musicnn-1.pb)..."
python -m tf2onnx.convert \
  --input ./mood_aggressive-msd-musicnn-1.pb \
  --inputs 'model/Placeholder:0' \
  --outputs 'model/Softmax:0' \
  --output mood_aggressive-msd-musicnn-1.onnx

echo "Converting happy mood model (mood_happy-msd-musicnn-1.pb)..."
python -m tf2onnx.convert \
  --input ./mood_happy-msd-musicnn-1.pb \
  --inputs 'model/Placeholder:0' \
  --outputs 'model/Softmax:0' \
  --output mood_happy-msd-musicnn-1.onnx

echo "Converting party mood model (mood_party-msd-musicnn-1.pb)..."
python -m tf2onnx.convert \
  --input ./mood_party-msd-musicnn-1.pb \
  --inputs 'model/Placeholder:0' \
  --outputs 'model/Softmax:0' \
  --output mood_party-msd-musicnn-1.onnx

echo "Converting relaxed mood model (mood_relaxed-msd-musicnn-1.pb)..."
python -m tf2onnx.convert \
  --input ./mood_relaxed-msd-musicnn-1.pb \
  --inputs 'model/Placeholder:0' \
  --outputs 'model/Softmax:0' \
  --output mood_relaxed-msd-musicnn-1.onnx

echo "Converting sad mood model (mood_sad-msd-musicnn-1.pb)..."
python -m tf2onnx.convert \
  --input ./mood_sad-msd-musicnn-1.pb \
  --inputs 'model/Placeholder:0' \
  --outputs 'model/Softmax:0' \
  --output mood_sad-msd-musicnn-1.onnx

echo "All models converted successfully!"

Before run the above converting command, a virtual envirorment was created and this dependencies was installed

pip install "numpy<2" tensorflow==2.13.1 tf2onnx==1.16.1 onnxruntime