Merge branch 'main' into feat/cuda

funasr/export/models/modules/multihead_att.py

@@ -75,8 +75,8 @@ def preprocess_for_attn(x, mask, cache, pad_fn):
     return x, cache
 
 
-torch_version = float(".".join(torch.__version__.split(".")[:2]))
-if torch_version >= 1.8:
+torch_version = tuple([int(i) for i in torch.__version__.split(".")[:2]])
+if torch_version >= (1, 8):
     import torch.fx
     torch.fx.wrap('preprocess_for_attn')
 

funasr/runtime/python/libtorch/demo.py

@@ -1,11 +1,15 @@
-
 from funasr_torch import Paraformer
 
+
 model_dir = "/nfs/zhifu.gzf/export/damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch"
+
 model = Paraformer(model_dir, batch_size=1)  # cpu
 # model = Paraformer(model_dir, batch_size=1, device_id=0)  # gpu
 
-wav_path = ['/nfs/zhifu.gzf/export/damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch/example/asr_example.wav']
+# when using paraformer-large-vad-punc model, you can set plot_timestamp_to="./xx.png" to get figure of alignment besides timestamps
+# model = Paraformer(model_dir, batch_size=1, plot_timestamp_to="test.png")
+
+wav_path = "YourPath/xx.wav"
 
 result = model(wav_path)
 print(result)

funasr/runtime/python/onnxruntime/demo.py

@@ -1,16 +1,15 @@
-
 from funasr_onnx import Paraformer
 
-#model_dir = "/Users/shixian/code/funasr/export/damo/speech_paraformer-large-vad-punc_asr_nat-zh-cn-16k-common-vocab8404-pytorch"
-#model_dir = "/Users/shixian/code/funasr/export/damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch"
-model_dir = "/Users/shixian/code/funasr/export/damo/speech_paraformer-tiny-commandword_asr_nat-zh-cn-16k-vocab544-pytorch"
+model_dir = "/nfs/zhifu.gzf/export/damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch"
 
-# if you use paraformer-tiny-commandword_asr_nat-zh-cn-16k-vocab544-pytorch, you should set pred_bias=0
-# plot_timestamp_to works only when using speech_paraformer-large-vad-punc_asr_nat-zh-cn-16k-common-vocab8404-pytorch
 model = Paraformer(model_dir, batch_size=2, plot_timestamp_to="./", pred_bias=0)  # cpu
 # model = Paraformer(model_dir, batch_size=2, plot_timestamp_to="./", pred_bias=0, device_id=0)  # gpu
 
-wav_path = "/Users/shixian/code/funasr/export/damo/speech_paraformer-tiny-commandword_asr_nat-zh-cn-16k-vocab544-pytorch/example/asr_example.wav"
+# when using paraformer-large-vad-punc model, you can set plot_timestamp_to="./xx.png" to get figure of alignment besides timestamps
+# model = Paraformer(model_dir, batch_size=1, plot_timestamp_to="test.png")
+
+
+wav_path = "YourPath/xx.wav"
 
 result = model(wav_path)
 print(result)