adapt view behavior change, fix KeyError. (PaddlePaddle#3794)

* adapt view behavior change, fix KeyError. * fix readme demo run error. * fixed opencc version
luotao1 · Jun 11, 2024 · bfeb8a0 · bfeb8a0
1 parent 2b7334e
commit bfeb8a0
Show file tree

Hide file tree

Showing 3 changed files with 3 additions and 3 deletions.
diff --git a/paddlespeech/cli/asr/infer.py b/paddlespeech/cli/asr/infer.py
@@ -274,7 +274,7 @@ def preprocess(self, model_type: str, input: Union[str, os.PathLike]):
             # fbank
             audio = preprocessing(audio, **preprocess_args)
 
-            audio_len = paddle.to_tensor([audio.shape[0]]).unsqueeze(axis=0)
+            audio_len = paddle.to_tensor(audio.shape[0]).unsqueeze(axis=0)
             audio = paddle.to_tensor(audio, dtype='float32').unsqueeze(axis=0)
 
             self._inputs["audio"] = audio

diff --git a/paddlespeech/s2t/models/wav2vec2/wav2vec2_ASR.py b/paddlespeech/s2t/models/wav2vec2/wav2vec2_ASR.py
@@ -188,7 +188,7 @@ def ctc_greedy_search(self, wav) -> List[List[int]]:
         x_lens = x.shape[1]
         ctc_probs = self.ctc.log_softmax(x)  # (B, maxlen, vocab_size)
         topk_prob, topk_index = ctc_probs.topk(1, axis=2)  # (B, maxlen, 1)
-        topk_index = topk_index.view([batch_size, x_lens])  # (B, maxlen)
+        topk_index = topk_index.reshape([batch_size, x_lens])  # (B, maxlen)
 
         hyps = [hyp.tolist() for hyp in topk_index]
         hyps = [remove_duplicates_and_blank(hyp) for hyp in hyps]

diff --git a/setup.py b/setup.py
@@ -48,7 +48,7 @@
     "matplotlib",
     "nara_wpe",
     "onnxruntime>=1.11.0",
-    "opencc",
+    "opencc==1.1.6",
     "opencc-python-reimplemented",
     "pandas",
     "paddleaudio>=1.1.0",