Feature Extraction
Transformers
Safetensors
moss-audio-tokenizer
audio
audio-tokenizer
neural-codec
moss-tts-family
MOSS Audio Tokenizer Nano
speech-tokenizer
trust-remote-code
custom_code
Instructions to use OpenMOSS-Team/MOSS-Audio-Tokenizer-Nano with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use OpenMOSS-Team/MOSS-Audio-Tokenizer-Nano with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("feature-extraction", model="OpenMOSS-Team/MOSS-Audio-Tokenizer-Nano", trust_remote_code=True)# Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("OpenMOSS-Team/MOSS-Audio-Tokenizer-Nano", trust_remote_code=True, dtype="auto") - Notebooks
- Google Colab
- Kaggle
Huge memory requirements
#2
by mllearner123 - opened
this function def _build_non_streaming_sdpa_bias() tries to take 4GB of memory even though the model is only 90MB.
at this line: delta = positions.view(1, max_seqlen, 1) - positions.view(1, 1, max_seqlen)
so this breaks GPUs with lower VRAM
maybe xformer masks could be applied or triton kernels