Automatic Speech Recognition
Transformers
Safetensors
PyTorch
arkasr
text-generation
speech
audio
vllm
ark-asr
custom_code
Eval Results
Instructions to use AutoArk-AI/ARK-ASR-3B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use AutoArk-AI/ARK-ASR-3B with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("automatic-speech-recognition", model="AutoArk-AI/ARK-ASR-3B", trust_remote_code=True)# Load model directly from transformers import AutoModelForCausalLM model = AutoModelForCausalLM.from_pretrained("AutoArk-AI/ARK-ASR-3B", trust_remote_code=True, device_map="auto") - Notebooks
- Google Colab
- Kaggle
there is no way to tell model LANG - its hallucinating
#4
by mikinko - opened
im testing it on europe lang set, as it has no tag for lang in audio , it hallucinate, mixing english with many languages, slavic langs are similar, so it has no chance, i did swap instructions to local lang , , put repetition_penalty 1.1 - it improved but beside main languages is useless.
Thanks a lot for your feedback and for reporting this issue!
We’re currently working on a new version that will support specifying the input language explicitly, which should significantly reduce language hallucination and mixing for multilingual inputs, especially for closely related languages. We’ll keep improving the multilingual performance. Thanks again for helping us improve the model!