Skip to content

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

Add model integration tests for major models (VAD, Paraformer, SenseVoice, CAM++, CT-Transformer, Streaming)

zhifu.gzf266b8a24aee8
diff --git a/tests_models/run_all_tests.py b/tests_models/run_all_tests.py
new file mode 100644
index 000000000..3b7ebe753
--- /dev/null
+++ b/tests_models/run_all_tests.py
@@ -0,0 +1,81 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""Run all model tests and report results"""
+import subprocess
+import sys
+import os
+import time
+
+TEST_DIR = os.path.dirname(os.path.abspath(__file__))
+PROJECT_DIR = os.path.dirname(TEST_DIR)
+
+tests = [
+ "test_fsmn_vad.py",
+ "test_ct_transformer.py",
+ "test_paraformer.py",
+ "test_sensevoice.py",
+ "test_campplus.py",
+ "test_paraformer_streaming.py",
+]
+
+SEP = "-" * 60
+DSEP = "=" * 60
+
+def main():
+ results = {}
+ total_start = time.time()
+
+ env = os.environ.copy()
+ env["PYTHONPATH"] = PROJECT_DIR + ":" + env.get("PYTHONPATH", "")
+
+ print(DSEP)
+ print("FunASR Model Tests")
+ print(DSEP)
+
+ for test_file in tests:
+ test_path = os.path.join(TEST_DIR, test_file)
+ print("\n" + SEP)
+ print("Running: " + test_file)
+ print(SEP)
+
+ t0 = time.time()
+ try:
+ result = subprocess.run(
+ [sys.executable, test_path],
+ cwd=PROJECT_DIR,
+ env=env,
+ timeout=300,
+ capture_output=False,
+ )
+ elapsed = time.time() - t0
+ results[test_file] = ("PASSED" if result.returncode == 0 else "FAILED", elapsed)
+ except subprocess.TimeoutExpired:
+ elapsed = time.time() - t0
+ results[test_file] = ("TIMEOUT", elapsed)
+ print(" TIMEOUT after %.1fs" % elapsed)
+ except Exception as e:
+ elapsed = time.time() - t0
+ results[test_file] = ("ERROR", elapsed)
+ print(" ERROR: %s" % e)
+
+ total_elapsed = time.time() - total_start
+
+ print("\n" + DSEP)
+ print("SUMMARY")
+ print(DSEP)
+ passed = 0
+ failed = 0
+ for test_file, (status, elapsed) in results.items():
+ icon = "+" if status == "PASSED" else "x"
+ print(" %s %-35s %-8s (%.1fs)" % (icon, test_file, status, elapsed))
+ if status == "PASSED":
+ passed += 1
+ else:
+ failed += 1
+
+ print("\n Total: %d passed, %d failed, %.1fs elapsed" % (passed, failed, total_elapsed))
+ print(DSEP)
+ return 0 if failed == 0 else 1
+
+if __name__ == "__main__":
+ sys.exit(main())
diff --git a/tests_models/test_campplus.py b/tests_models/test_campplus.py
new file mode 100644
index 000000000..1bf26a2c7
--- /dev/null
+++ b/tests_models/test_campplus.py
@@ -0,0 +1,31 @@
+#!/usr/bin/env python3
+"""Test CAM++: Speaker Verification"""
+import sys
+import time
+
+def main():
+ from funasr import AutoModel
+
+ print("[CAM++] Loading model...")
+ t0 = time.time()
+ model = AutoModel(model="iic/speech_campplus_sv_zh-cn_16k-common", device="cpu", disable_update=True)
+ print("[CAM++] Model loaded in %.1fs" % (time.time()-t0))
+
+ print("[CAM++] Extracting speaker embedding...")
+ t0 = time.time()
+ res = model.generate(
+ input="https://isv-data.oss-cn-hangzhou.aliyuncs.com/ics/MaaS/ASR/test_audio/asr_example_zh.wav",
+ )
+ print("[CAM++] Inference done in %.1fs" % (time.time()-t0))
+
+ if res and len(res) > 0:
+ spk_embedding = res[0].get("spk_embedding", None)
+ if spk_embedding is not None:
+ print("[CAM++] Embedding shape: %s" % str(spk_embedding.shape))
+ print("[CAM++] PASSED")
+ return 0
+ print("[CAM++] FAILED - no speaker embedding")
+ return 1
+
+if __name__ == "__main__":
+ sys.exit(main())
diff --git a/tests_models/test_ct_transformer.py b/tests_models/test_ct_transformer.py
new file mode 100644
index 000000000..40d9ae261
--- /dev/null
+++ b/tests_models/test_ct_transformer.py
@@ -0,0 +1,28 @@
+#!/usr/bin/env python3
+"""Test CT-Transformer: Punctuation Restoration"""
+import sys
+import time
+
+def main():
+ from funasr import AutoModel
+
+ print("[CT-Transformer] Loading model...")
+ t0 = time.time()
+ model = AutoModel(model="iic/punc_ct-transformer_zh-cn-common-vocab272727-pytorch", device="cpu", disable_update=True)
+ print("[CT-Transformer] Model loaded in %.1fs" % (time.time()-t0))
+
+ print("[CT-Transformer] Running inference...")
+ t0 = time.time()
+ res = model.generate(input="那今天的天气呢也是蛮好的啊你觉得怎么样呢我觉得还不错")
+ print("[CT-Transformer] Inference done in %.1fs" % (time.time()-t0))
+ print("[CT-Transformer] Result: %s" % res)
+
+ if res and len(res) > 0 and "text" in res[0]:
+ print("[CT-Transformer] PASSED")
+ return 0
+ else:
+ print("[CT-Transformer] FAILED - no punctuation result")
+ return 1
+
+if __name__ == "__main__":
+ sys.exit(main())
diff --git a/tests_models/test_fsmn_vad.py b/tests_models/test_fsmn_vad.py
new file mode 100644
index 000000000..28a09553c
--- /dev/null
+++ b/tests_models/test_fsmn_vad.py
@@ -0,0 +1,31 @@
+#!/usr/bin/env python3
+"""Test FSMN-VAD: Voice Activity Detection"""
+import sys
+import time
+
+def main():
+ from funasr import AutoModel
+
+ print("[FSMN-VAD] Loading model...")
+ t0 = time.time()
+ model = AutoModel(model="iic/speech_fsmn_vad_zh-cn-16k-common-pytorch", device="cpu", disable_update=True)
+ print("[FSMN-VAD] Model loaded in %.1fs" % (time.time()-t0))
+
+ print("[FSMN-VAD] Running inference...")
+ t0 = time.time()
+ res = model.generate(
+ input="https://isv-data.oss-cn-hangzhou.aliyuncs.com/ics/MaaS/ASR/test_audio/asr_example_zh.wav",
+ )
+ print("[FSMN-VAD] Inference done in %.1fs" % (time.time()-t0))
+ print("[FSMN-VAD] Result: %s" % res)
+
+ if res and len(res) > 0 and "value" in res[0]:
+ print("[FSMN-VAD] Detected %d speech segments" % len(res[0]["value"]))
+ print("[FSMN-VAD] PASSED")
+ return 0
+ else:
+ print("[FSMN-VAD] FAILED - no VAD segments detected")
+ return 1
+
+if __name__ == "__main__":
+ sys.exit(main())
diff --git a/tests_models/test_paraformer.py b/tests_models/test_paraformer.py
new file mode 100644
index 000000000..38afcde37
--- /dev/null
+++ b/tests_models/test_paraformer.py
@@ -0,0 +1,38 @@
+#!/usr/bin/env python3
+"""Test Paraformer-large: offline ASR (Chinese)"""
+import sys
+import time
+
+def main():
+ from funasr import AutoModel
+
+ print("[Paraformer] Loading model...")
+ t0 = time.time()
+ model = AutoModel(
+ model="iic/speech_paraformer-large-vad-punc_asr_nat-zh-cn-16k-common-vocab8404-pytorch",
+ vad_model="iic/speech_fsmn_vad_zh-cn-16k-common-pytorch",
+ vad_kwargs={"max_single_segment_time": 60000},
+ punc_model="iic/punc_ct-transformer_zh-cn-common-vocab272727-pytorch",
+ device="cpu",
+ disable_update=True,
+ )
+ print("[Paraformer] Model loaded in %.1fs" % (time.time()-t0))
+
+ print("[Paraformer] Running inference...")
+ t0 = time.time()
+ res = model.generate(
+ input="https://isv-data.oss-cn-hangzhou.aliyuncs.com/ics/MaaS/ASR/test_audio/asr_example_zh.wav",
+ cache={},
+ )
+ print("[Paraformer] Inference done in %.1fs" % (time.time()-t0))
+ print("[Paraformer] Result: %s" % res)
+
+ if res and len(res) > 0 and "text" in res[0]:
+ print("[Paraformer] PASSED")
+ return 0
+ else:
+ print("[Paraformer] FAILED - no text in result")
+ return 1
+
+if __name__ == "__main__":
+ sys.exit(main())
diff --git a/tests_models/test_paraformer_streaming.py b/tests_models/test_paraformer_streaming.py
new file mode 100644
index 000000000..99aec136e
--- /dev/null
+++ b/tests_models/test_paraformer_streaming.py
@@ -0,0 +1,33 @@
+#!/usr/bin/env python3
+"""Test Paraformer-Streaming: online/streaming ASR"""
+import sys
+import time
+
+def main():
+ from funasr import AutoModel
+
+ print("[Paraformer-Streaming] Loading model...")
+ t0 = time.time()
+ model = AutoModel(
+ model="iic/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-online",
+ device="cpu",
+ disable_update=True,
+ )
+ print("[Paraformer-Streaming] Model loaded in %.1fs" % (time.time()-t0))
+
+ print("[Paraformer-Streaming] Running inference...")
+ t0 = time.time()
+ test_wav = "https://isv-data.oss-cn-hangzhou.aliyuncs.com/ics/MaaS/ASR/test_audio/asr_example_zh.wav"
+ res = model.generate(input=test_wav, cache={}, is_final=True)
+ print("[Paraformer-Streaming] Inference done in %.1fs" % (time.time()-t0))
+ print("[Paraformer-Streaming] Result: %s" % res)
+
+ if res and len(res) > 0 and "text" in res[0]:
+ print("[Paraformer-Streaming] PASSED")
+ return 0
+ else:
+ print("[Paraformer-Streaming] FAILED - no text in result")
+ return 1
+
+if __name__ == "__main__":
+ sys.exit(main())
diff --git a/tests_models/test_sensevoice.py b/tests_models/test_sensevoice.py
new file mode 100644
index 000000000..a9c0df0b4
--- /dev/null
+++ b/tests_models/test_sensevoice.py
@@ -0,0 +1,44 @@
+#!/usr/bin/env python3
+"""Test SenseVoice: multi-task speech understanding"""
+import sys
+import time
+
+def main():
+ from funasr import AutoModel
+ from funasr.utils.postprocess_utils import rich_transcription_postprocess
+
+ print("[SenseVoice] Loading model...")
+ t0 = time.time()
+ model = AutoModel(
+ model="iic/SenseVoiceSmall",
+ vad_model="fsmn-vad",
+ vad_kwargs={"max_single_segment_time": 30000},
+ device="cpu",
+ disable_update=True,
+ )
+ print("[SenseVoice] Model loaded in %.1fs" % (time.time()-t0))
+
+ print("[SenseVoice] Running inference (Chinese)...")
+ t0 = time.time()
+ res = model.generate(
+ input=model.model_path + "/example/zh.mp3",
+ cache={},
+ language="auto",
+ use_itn=True,
+ batch_size_s=60,
+ merge_vad=True,
+ merge_length_s=15,
+ )
+ print("[SenseVoice] Inference done in %.1fs" % (time.time()-t0))
+
+ if res and len(res) > 0 and "text" in res[0]:
+ text = rich_transcription_postprocess(res[0]["text"])
+ print("[SenseVoice] Result: %s" % text)
+ print("[SenseVoice] PASSED")
+ return 0
+ else:
+ print("[SenseVoice] FAILED - no text in result")
+ return 1
+
+if __name__ == "__main__":
+ sys.exit(main())