Skip to content

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

fix: vLLM >=0.21 compat — EmbedsPrompt moved from vllm.inputs.data to vllm.inputs

游雁39530728fec2
diff --git a/funasr/models/fun_asr_nano/inference_vllm.py b/funasr/models/fun_asr_nano/inference_vllm.py
index 38384b509..e6cb95aeb 100644
--- a/funasr/models/fun_asr_nano/inference_vllm.py
+++ b/funasr/models/fun_asr_nano/inference_vllm.py
@@ -167,7 +167,10 @@ class FunASRNanoVLLM:
**kwargs,
):
from vllm import LLM, SamplingParams
- from vllm.inputs.data import EmbedsPrompt
+ try:
+ from vllm.inputs import EmbedsPrompt
+ except ImportError:
+ from vllm.inputs.data import EmbedsPrompt
self.device = device
self.dtype = dtype
@@ -519,7 +522,10 @@ class FunASRNanoVLLM:
List of result dicts: [{"key": str, "text": str, "timestamps": [...]}]
"""
from vllm import SamplingParams
- from vllm.inputs.data import EmbedsPrompt
+ try:
+ from vllm.inputs import EmbedsPrompt
+ except ImportError:
+ from vllm.inputs.data import EmbedsPrompt
if isinstance(inputs, (str, np.ndarray, torch.Tensor)):
inputs = [inputs]
diff --git a/funasr/models/fun_asr_nano/inference_vllm_pipeline.py b/funasr/models/fun_asr_nano/inference_vllm_pipeline.py
index d7afc2f6d..bef4cb7a2 100644
--- a/funasr/models/fun_asr_nano/inference_vllm_pipeline.py
+++ b/funasr/models/fun_asr_nano/inference_vllm_pipeline.py
@@ -199,7 +199,10 @@ class FunASRNanoVLLMPipeline:
# Step 3: Batch ASR via vLLM
# Encode all segments and build prompts
from vllm import SamplingParams
- from vllm.inputs.data import EmbedsPrompt
+ try:
+ from vllm.inputs import EmbedsPrompt
+ except ImportError:
+ from vllm.inputs.data import EmbedsPrompt
prompts = []
for seg_audio in segment_audios:
diff --git a/funasr/models/fun_asr_nano/inference_vllm_streaming.py b/funasr/models/fun_asr_nano/inference_vllm_streaming.py
index cdc17ebf1..4ff1f9ccf 100644
--- a/funasr/models/fun_asr_nano/inference_vllm_streaming.py
+++ b/funasr/models/fun_asr_nano/inference_vllm_streaming.py
@@ -210,7 +210,10 @@ class FunASRNanoStreamingVLLM:
"is_final": bool, "chunk_idx": int, "audio_duration_ms": float}
"""
from vllm import SamplingParams
- from vllm.inputs.data import EmbedsPrompt
+ try:
+ from vllm.inputs import EmbedsPrompt
+ except ImportError:
+ from vllm.inputs.data import EmbedsPrompt
from funasr.utils.load_utils import load_audio_text_image_video
chunk_ms = chunk_ms or self.chunk_ms
diff --git a/funasr/models/glm_asr/inference_vllm.py b/funasr/models/glm_asr/inference_vllm.py
index d2eaac625..73687a335 100644
--- a/funasr/models/glm_asr/inference_vllm.py
+++ b/funasr/models/glm_asr/inference_vllm.py
@@ -192,7 +192,10 @@ class GLMASRVLLMEngine:
List of {"key": str, "text": str}
"""
from vllm import SamplingParams
- from vllm.inputs.data import EmbedsPrompt
+ try:
+ from vllm.inputs import EmbedsPrompt
+ except ImportError:
+ from vllm.inputs.data import EmbedsPrompt
if isinstance(inputs, (str, np.ndarray, torch.Tensor)):
inputs = [inputs]
diff --git a/funasr/version.txt b/funasr/version.txt
index d0149fef7..80e78df68 100644
--- a/funasr/version.txt
+++ b/funasr/version.txt
@@ -1 +1 @@
-1.3.4
+1.3.5