Skip to content

Open-source speech recognition toolkit for training, inference, streaming ASR, VAD, punctuation, speaker diarization pipelines, and OpenAI-compatible/MCP serving.

docs: fix vLLM version requirement, add AutoModelVLLM usage to vllm_inference.md

xiaoyunchong.xyc55c035ff4d17
diff --git a/examples/industrial_data_pretraining/fun_asr_nano/docs/vllm_inference.md b/examples/industrial_data_pretraining/fun_asr_nano/docs/vllm_inference.md
index c3b39ec44..f4860d648 100644
--- a/examples/industrial_data_pretraining/fun_asr_nano/docs/vllm_inference.md
+++ b/examples/industrial_data_pretraining/fun_asr_nano/docs/vllm_inference.md
@@ -15,7 +15,7 @@ This architecture allows vLLM to efficiently batch multiple requests and leverag
```bash
# Core dependencies
pip install funasr>=1.3.0
-pip install vllm>=0.11.0
+pip install vllm>=0.12.0
pip install safetensors
pip install tiktoken
@@ -25,7 +25,23 @@ pip install modelscope
## Quick Start
-### Python API
+### AutoModelVLLM (Recommended)
+
+```python
+from funasr.auto.auto_model_vllm import AutoModelVLLM
+
+# Generic interface - auto-detects model type, works for all LLM-based ASR models
+model = AutoModelVLLM(
+ model="FunAudioLLM/Fun-ASR-Nano-2512",
+ hub="ms",
+ tensor_parallel_size=1,
+)
+
+results = model.generate(["audio.wav"], language="中文")
+print(results[0]["text"])
+```
+
+### FunASRNanoVLLM (Direct API)
```python
from funasr.models.fun_asr_nano.inference_vllm import FunASRNanoVLLM