Automatic Speech Recognition
Transformers
Safetensors
fun_asr_nano
speech-recognition
asr
end-to-end
multilingual
arxiv:2509.12508
Instructions to use FunAudioLLM/Fun-ASR-Nano-2512-hf with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use FunAudioLLM/Fun-ASR-Nano-2512-hf with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("automatic-speech-recognition", model="FunAudioLLM/Fun-ASR-Nano-2512-hf")# pip install -U transformers accelerate # Load model directly from transformers import AutoProcessor, AutoModelForMultimodalLM processor = AutoProcessor.from_pretrained("FunAudioLLM/Fun-ASR-Nano-2512-hf") model = AutoModelForMultimodalLM.from_pretrained("FunAudioLLM/Fun-ASR-Nano-2512-hf", device_map="auto") - Notebooks
- Google Colab
- Kaggle
Commit History
-
Regenerate checkpoint for current Transformers layout
Verified signatureRegenerated from the original Fun-ASR-Nano ModelScope checkpoint after the Whisper component refactor. Local conversion loaded 1,540 tensors with zero missing, unexpected, or mismatched keys; H100 Chinese, English, default-prompt, and batch inference matched expected outputs.c49669b
-
Refresh checkpoint for shared Whisper component layout
Verified signatureRe-converted from source revision 272c57b with Transformers PR #46180 after the shared audio-component refactor. Local validation: 1540 tensors, zero missing/unexpected/mismatched keys; exact Chinese, English, and batch H100 outputs; PR RUN_SLOW integration tests 3 passed.6367f42
-
docs: add tested Transformers PR install command
zhifu gaob8b7ac6
-
docs: clarify Nano language coverage (#1)
- docs: clarify Nano language coverage (6715224cf0a749bc092b2dec9d76057537e3f630)e711f3d
-
Update split-layout Transformers checkpoint
Verified signaturePublish the checkpoint and processor artifacts validated for huggingface/transformers#46180. The model loads with zero missing, unexpected, or mismatched keys and passes the H100 English transcription quickstart. Also replace the stale preprocessor config and add model-card usage instructions.4d5668e
-
Re-convert Fun-ASR-Nano to dtype-consistent bf16 (modular split layout); shrink 2.75GB->1.97GB
Verified signature35bf550 -
Re-convert Fun-ASR-Nano to dtype-consistent bf16 (modular split layout); shrink 2.75GB->1.97GB
Verified signature2aa316d
-
Add chat_template.jinja for transformers apply_chat_template
Verified signaturef2ba1bd -
docs: add FunASR ecosystem GitHub Star links
3ecd9e9
-
Add README with pipeline_tag: automatic-speech-recognition
Zhifu Gao9436970
-
Add Fun-ASR-Nano in HuggingFace Transformers format (safetensors)
Verified signature1b829ba -
initial commit
Verified signatureb405a1c