yuxinlu1/gemma-4-12B-coder-fable5-composer2.5-v1-GGUF
Text Generation ā¢ 12B ā¢ Updated ā¢ 483k ā¢ 2.31k
Matteo Spanio
matteospanio
AI & ML interests
Generative AI, AI for audio restoration, Generative music systems
Recent Activity
liked a model 8 days ago
yuxinlu1/gemma-4-12B-coder-fable5-composer2.5-v1-GGUF upvoted a changelog 17 days ago
Publish models from CI without HF_TOKEN reacted to theirpost with š 18 days ago
š¶ Released mule-torch ā an unofficial PyTorch port of MULE (SF-NFNet-F0), SiriusXM/Pandora's music-audio embedding model (McCallum et al., ISMIR 2022).
No retraining: I re-implemented the architecture in pure PyTorch and transferred the original TensorFlow weights, then checked it layer by layer against the genuine TF pipeline.
ā
End-to-end clip-embedding cosine 0.9999999 vs the original
ā
ONNX backbone parity < 1e-6
ā
62.35M params (paper: ~62.4M)
ā
Batched, GPU-native, ONNX-exportable ā none of which the original `Analysis` pipeline does
```python
pip install mule-torch
```
```python
from mule_torch import MuleModel
emb = MuleModel.from_pretrained()(waveform) # (B, T)@16kHz -> (B, 1728)
```
š¤ Weights: https://huggingface.co/matteospanio/mule
š» Code: https://github.com/matteospanio/mule-torch
š¦ PyPI: https://pypi.org/project/mule-torch/
The fun bug: parity was perfect through every conv but the block output was anti-correlated (cos = ā1). Cause: the learnable skip-init gains couldn't be mapped by layer name (Keras scrambles the order) ā they had to be recovered from the graph.
ā ļø Unofficial, community port ā not affiliated with or endorsed by the original authors. All credit to them; please cite the paper. Weights inherit CC-BY-NC-4.0.Organizations
liked a
model 8 days ago
upvoted a changelog 17 days ago
Hugging Face Changelog
Publish models from CI without HF_TOKEN
ā¢ 105
reacted to their post with š 18 days ago
Post
7052
š¶ Released mule-torch ā an unofficial PyTorch port of MULE (SF-NFNet-F0), SiriusXM/Pandora's music-audio embedding model (McCallum et al., ISMIR 2022).
No retraining: I re-implemented the architecture in pure PyTorch and transferred the original TensorFlow weights, then checked it layer by layer against the genuine TF pipeline.
ā End-to-end clip-embedding cosine 0.9999999 vs the original
ā ONNX backbone parity < 1e-6
ā 62.35M params (paper: ~62.4M)
ā Batched, GPU-native, ONNX-exportable ā none of which the original
š¤ Weights: matteospanio/mule
š» Code: https://github.com/matteospanio/mule-torch
š¦ PyPI: https://pypi.org/project/mule-torch/
The fun bug: parity was perfect through every conv but the block output was anti-correlated (cos = ā1). Cause: the learnable skip-init gains couldn't be mapped by layer name (Keras scrambles the order) ā they had to be recovered from the graph.
ā ļø Unofficial, community port ā not affiliated with or endorsed by the original authors. All credit to them; please cite the paper. Weights inherit CC-BY-NC-4.0.
No retraining: I re-implemented the architecture in pure PyTorch and transferred the original TensorFlow weights, then checked it layer by layer against the genuine TF pipeline.
ā End-to-end clip-embedding cosine 0.9999999 vs the original
ā ONNX backbone parity < 1e-6
ā 62.35M params (paper: ~62.4M)
ā Batched, GPU-native, ONNX-exportable ā none of which the original
Analysis pipeline doespip install mule-torchfrom mule_torch import MuleModel
emb = MuleModel.from_pretrained()(waveform) # (B, T)@16kHz -> (B, 1728)š¤ Weights: matteospanio/mule
š» Code: https://github.com/matteospanio/mule-torch
š¦ PyPI: https://pypi.org/project/mule-torch/
The fun bug: parity was perfect through every conv but the block output was anti-correlated (cos = ā1). Cause: the learnable skip-init gains couldn't be mapped by layer name (Keras scrambles the order) ā they had to be recovered from the graph.
ā ļø Unofficial, community port ā not affiliated with or endorsed by the original authors. All credit to them; please cite the paper. Weights inherit CC-BY-NC-4.0.
liked a
model 18 days ago
posted an update 18 days ago
Post
7052
š¶ Released mule-torch ā an unofficial PyTorch port of MULE (SF-NFNet-F0), SiriusXM/Pandora's music-audio embedding model (McCallum et al., ISMIR 2022).
No retraining: I re-implemented the architecture in pure PyTorch and transferred the original TensorFlow weights, then checked it layer by layer against the genuine TF pipeline.
ā End-to-end clip-embedding cosine 0.9999999 vs the original
ā ONNX backbone parity < 1e-6
ā 62.35M params (paper: ~62.4M)
ā Batched, GPU-native, ONNX-exportable ā none of which the original
š¤ Weights: matteospanio/mule
š» Code: https://github.com/matteospanio/mule-torch
š¦ PyPI: https://pypi.org/project/mule-torch/
The fun bug: parity was perfect through every conv but the block output was anti-correlated (cos = ā1). Cause: the learnable skip-init gains couldn't be mapped by layer name (Keras scrambles the order) ā they had to be recovered from the graph.
ā ļø Unofficial, community port ā not affiliated with or endorsed by the original authors. All credit to them; please cite the paper. Weights inherit CC-BY-NC-4.0.
No retraining: I re-implemented the architecture in pure PyTorch and transferred the original TensorFlow weights, then checked it layer by layer against the genuine TF pipeline.
ā End-to-end clip-embedding cosine 0.9999999 vs the original
ā ONNX backbone parity < 1e-6
ā 62.35M params (paper: ~62.4M)
ā Batched, GPU-native, ONNX-exportable ā none of which the original
Analysis pipeline doespip install mule-torchfrom mule_torch import MuleModel
emb = MuleModel.from_pretrained()(waveform) # (B, T)@16kHz -> (B, 1728)š¤ Weights: matteospanio/mule
š» Code: https://github.com/matteospanio/mule-torch
š¦ PyPI: https://pypi.org/project/mule-torch/
The fun bug: parity was perfect through every conv but the block output was anti-correlated (cos = ā1). Cause: the learnable skip-init gains couldn't be mapped by layer name (Keras scrambles the order) ā they had to be recovered from the graph.
ā ļø Unofficial, community port ā not affiliated with or endorsed by the original authors. All credit to them; please cite the paper. Weights inherit CC-BY-NC-4.0.
updated a
model 19 days ago
published a
model 19 days ago
upvoted a paper 30 days ago
liked a
Space 30 days ago
liked a
model about 1 month ago
authored a
paper 2 months ago
upvoted a paper 2 months ago
liked a
model 2 months ago
upvoted a paper 2 months ago
submitted a
paper to Daily Papers 2 months ago
updated a
model 2 months ago
liked a
model 3 months ago
published a
model 3 months ago
liked a
dataset 3 months ago
posted an update 4 months ago
Post
156
š§ Help us evaluate AI-generated music across cultures
We are running a new online survey to collect data for a follow-up study on AI-generated music and sonic seasoning: the phenomenon where sound can influence the perception of taste.
This study builds on our previous work:
š Paper: A Multimodal Symphony: Integrating Taste and Sound through Generative AI (2503.02823)
š¤ Model release: csc-unipd/tasty-musicgen-small
In the previous study we introduced Tasty MusicGen, a model designed to generate music associated with taste descriptors, and showed that it can effectively produce music that evokes specific taste-related sensations.
With this new survey, we aim to expand the participant pool to create a more inclusive and cross-cultural evaluation, helping us understand how these musical cues are perceived across different linguistic and cultural backgrounds.
Participants will listen to short AI-generated music clips and evaluate the sensations they evoke.
ā±ļø Takes about 10 minutes
š Participate here:
https://matteospanio.github.io/tasty-music-survey/
Your participation will help us better understand how generative music models interact with human perception across cultures. Thanks to everyone who participates or shares!
We are running a new online survey to collect data for a follow-up study on AI-generated music and sonic seasoning: the phenomenon where sound can influence the perception of taste.
This study builds on our previous work:
š Paper: A Multimodal Symphony: Integrating Taste and Sound through Generative AI (2503.02823)
š¤ Model release: csc-unipd/tasty-musicgen-small
In the previous study we introduced Tasty MusicGen, a model designed to generate music associated with taste descriptors, and showed that it can effectively produce music that evokes specific taste-related sensations.
With this new survey, we aim to expand the participant pool to create a more inclusive and cross-cultural evaluation, helping us understand how these musical cues are perceived across different linguistic and cultural backgrounds.
Participants will listen to short AI-generated music clips and evaluate the sensations they evoke.
ā±ļø Takes about 10 minutes
š Participate here:
https://matteospanio.github.io/tasty-music-survey/
Your participation will help us better understand how generative music models interact with human perception across cultures. Thanks to everyone who participates or shares!