diff --git a/README.md b/README.md index d10fd5d59..de24abe2f 100644 --- a/README.md +++ b/README.md @@ -15,6 +15,7 @@ ## Setup +All tested under: * Ubuntu 16.04 * python>=3.7 * paddlepaddle>=2.1.2 diff --git a/README_cn.md b/README_cn.md index 90a65c440..29aadbdf6 100644 --- a/README_cn.md +++ b/README_cn.md @@ -16,6 +16,8 @@ ## 安装 +在以下环境测试验证过: + * Ubuntu 16.04 * python>=3.7 * paddlepaddle>=2.1.2 diff --git a/deepspeech/exps/u2_kaldi/model.py b/deepspeech/exps/u2_kaldi/model.py index 60f070a3b..a2f062a18 100644 --- a/deepspeech/exps/u2_kaldi/model.py +++ b/deepspeech/exps/u2_kaldi/model.py @@ -233,7 +233,7 @@ class U2Trainer(Trainer): batch_frames_inout=0, preprocess_conf=config.collator.augmentation_config, n_iter_processes=config.collator.num_workers, - subsampling_factor=1, + subsampling_factor=0, num_encs=1) self.valid_loader = BatchDataLoader( diff --git a/deepspeech/io/converter.py b/deepspeech/io/converter.py index a02e06acb..e591a7935 100644 --- a/deepspeech/io/converter.py +++ b/deepspeech/io/converter.py @@ -55,6 +55,8 @@ class CustomConverter(): xs = [x[::self.subsampling_factor, :] for x in xs] # get batch of lengths of input sequences + print(xs) + print(ys) ilens = np.array([x.shape[0] for x in xs]) # perform padding and convert to tensor