mp1704/vits_tram_radio_6h40
83M β’ Updated β’ 3
k0 c@n gj@j th!ch π
mp1704
AI & ML interests
None yet
Recent Activity
reacted to matteospanio's post with π about 12 hours ago
πΆ Released mule-torch β 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.
β
End-to-end clip-embedding cosine 0.9999999 vs the original
β
ONNX backbone parity < 1e-6
β
62.35M params (paper: ~62.4M)
β
Batched, GPU-native, ONNX-exportable β 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)
```
π€ Weights: https://huggingface.co/matteospanio/mule
π» Code: https://github.com/matteospanio/mule-torch
π¦ PyPI: https://pypi.org/project/mule-torch/
The fun bug: parity was perfect through every conv but the block output was anti-correlated (cos = β1). Cause: the learnable skip-init gains couldn't be mapped by layer name (Keras scrambles the order) β they had to be recovered from the graph.
β οΈ Unofficial, community port β 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 dataset about 16 hours ago
mp1704/onnx-weight-08-june published a dataset about 16 hours ago
mp1704/onnx-weight-08-juneOrganizations
mp1704/huhu
Updated
mp1704/adapter_65h_10speakers
Updated β’ 1
mp1704/csm-1b-20h-ckpt1000
Text-to-Audio β’ 2B β’ Updated
mp1704/tora_7b_sft_ckpt_200
Text Generation β’ 7B β’ Updated β’ 2
mp1704/tora_7b_pt
Text Generation β’ 7B β’ Updated β’ 3
mp1704/gpt-neo-sft-v2.1
Text Generation β’ 0.4B β’ Updated β’ 1
mp1704/gpt-neo-sft-v2
Text Generation β’ 0.4B β’ Updated β’ 2
mp1704/gpt-neo-sft
Text Generation β’ 0.4B β’ Updated β’ 2
mp1704/gpt-neo-pt
Text Generation β’ 0.4B β’ Updated β’ 5
mp1704/gemma_2b_sft
Text Generation β’ 3B β’ Updated β’ 6
mp1704/gemma_2b_pt
Text Generation β’ 3B β’ Updated β’ 3
mp1704/qwen_1.8b_sft_full_3
Text Generation β’ 2B β’ Updated β’ 2 β’
mp1704/qwen_1.8b_sft_full_2
Feature Extraction β’ 2B β’ Updated β’ 2
mp1704/qwen_1.8b_sft_full_1
Feature Extraction β’ 2B β’ Updated β’ 2
mp1704/qwen_1.8b_sft_full
Text Generation β’ Updated β’ 1
mp1704/qwen_1.8b_stage_2
Text Generation β’ Updated β’ 6
mp1704/demo
Updated
mp1704/qwen_1.8b_stage_1
Text Generation β’ Updated β’ 2