Skip to content

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

update scripts

R1ckShi0f1247d7a8d4
diff --git a/examples/industrial_data_pretraining/bicif_paraformer/export.py b/examples/industrial_data_pretraining/bicif_paraformer/export.py
index 31098d2fd..44849b066 100644
--- a/examples/industrial_data_pretraining/bicif_paraformer/export.py
+++ b/examples/industrial_data_pretraining/bicif_paraformer/export.py
@@ -12,17 +12,17 @@ model = AutoModel(
device="cpu",
)
-res = model.export(type="onnx", quantize=False)
+res = model.export(type="torchscripts", quantize=False)
print(res)
-# method2, inference from local path
-from funasr import AutoModel
+# # method2, inference from local path
+# from funasr import AutoModel
-model = AutoModel(
- model="/Users/zhifu/.cache/modelscope/hub/iic/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch",
- device="cpu",
-)
+# model = AutoModel(
+# model="/Users/zhifu/.cache/modelscope/hub/iic/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch",
+# device="cpu",
+# )
-res = model.export(type="onnx", quantize=False)
-print(res)
+# res = model.export(type="onnx", quantize=False)
+# print(res)
diff --git a/funasr/utils/export_utils.py b/funasr/utils/export_utils.py
index 8f1aa5390..d205400c1 100644
--- a/funasr/utils/export_utils.py
+++ b/funasr/utils/export_utils.py
@@ -22,6 +22,7 @@ def export(model, data_in=None, quantize: bool = False, opset_version: int = 14,
)
elif type == 'torchscripts':
device = 'cuda' if torch.cuda.is_available() else 'cpu'
+ print("Exporting torchscripts on device {}".format(device))
_torchscripts(
m,
path=export_dir,
diff --git a/runtime/python/libtorch/funasr_torch/utils/timestamp_utils.py b/runtime/python/libtorch/funasr_torch/utils/timestamp_utils.py
index 5abbafed8..a10d19387 100644
--- a/runtime/python/libtorch/funasr_torch/utils/timestamp_utils.py
+++ b/runtime/python/libtorch/funasr_torch/utils/timestamp_utils.py
@@ -7,7 +7,7 @@ def time_stamp_lfr6_onnx(us_cif_peak, char_list, begin_time=0.0, total_offset=-1
START_END_THRESHOLD = 5
MAX_TOKEN_DURATION = 30
TIME_RATE = 10.0 * 6 / 1000 / 3 # 3 times upsampled
- cif_peak = us_cif_peak.reshape(-1)
+ cif_peak = us_cif_peak.reshape(-1).cpu()
num_frames = cif_peak.shape[-1]
if char_list[-1] == "</s>":
char_list = char_list[:-1]