Any-to-Any
Transformers
Safetensors
English
step_audio_2
text-generation
costa
speech-to-speech
spoken-dialogue
speech-token-compression
custom_code
Instructions to use D-YZ/CoSTA-EN with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use D-YZ/CoSTA-EN with Transformers:
# pip install -U transformers accelerate # Load model directly from transformers import AutoModelForCausalLM model = AutoModelForCausalLM.from_pretrained("D-YZ/CoSTA-EN", trust_remote_code=True, device_map="auto") - Notebooks
- Google Colab
- Kaggle
Download NOTICE from D-YZ/CoSTA-EN: direct link, hf CLI and curl.
- Browser
- Download file 550 Bytes
-
https://huggingface.co/D-YZ/CoSTA-EN/resolve/main/NOTICE
- Command line
-
hf download hf://D-YZ/CoSTA-EN/NOTICE
-
curl -L -o NOTICE https://huggingface.co/D-YZ/CoSTA-EN/resolve/main/NOTICE
550 Bytes
| CoSTA-EN | |
| CoSTA: Compressed Speech Token Adaptation for Efficient Spoken Dialogue Models | |
| https://arxiv.org/abs/2610.21143 | |
| https://github.com/yingzhuodeng/CoSTA | |
| Based on Step-Audio-2-mini by StepFun: | |
| https://huggingface.co/stepfun-ai/Step-Audio-2-mini | |
| The upstream model and implementation are licensed under Apache-2.0. | |
| CoSTA adds compressed speech-token vocabulary adaptation and dialogue fine-tuning. | |
| The custom model code includes changes for training loss, embedding/LM-head | |
| resizing, and inference timing. Existing source notices are retained. | |