Automatic Speech Recognition
Transformers
Safetensors
Danish
cohere_asr
audio
speech-recognition
transcription
danish
hf-asr-leaderboard
custom_code
Instructions to use syvai/hviske-v5.3 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use syvai/hviske-v5.3 with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("automatic-speech-recognition", model="syvai/hviske-v5.3", trust_remote_code=True)# Load model directly from transformers import AutoProcessor, AutoModelForSpeechSeq2Seq processor = AutoProcessor.from_pretrained("syvai/hviske-v5.3", trust_remote_code=True) model = AutoModelForSpeechSeq2Seq.from_pretrained("syvai/hviske-v5.3", trust_remote_code=True, device_map="auto") - Notebooks
- Google Colab
- Kaggle
GGUF - or help getting there?
#5
by GeoMaciolek - opened
Hey - I've been hoping to use your model with CrispASR (to use it to generate subtitles for Den Store Strikkedyst!) which works with GGUF files.
However, when trying to convert to GGUF using CrispASR's convert-cohere-asr-to-gguf.py, I error out as we're missing these (layers? tensors?):
preprocessor.featurizer.fb[1, 128, 257]preprocessor.featurizer.window[400]
(I modified the convert...py script to skip those, and the inference engine in CrispASR doesn't work, complains thus (these are the corresponding tensors to the ones above):
cohere: tensor 'fe.mel_fb' not found in GGUF
cohere: tensor 'fe.window' not found in GGUF
cohere: BN folded into conv_dw weights for 48 layers