From 8aa94b9254518417cd19059c4411168ec3488358 Mon Sep 17 00:00:00 2001 From: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Date: Wed, 9 Aug 2023 12:48:30 -0700 Subject: [PATCH] minor fix for conformer subsampling docstring. (#7195) Signed-off-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Signed-off-by: dorotat --- nemo/collections/asr/modules/conformer_encoder.py | 2 +- nemo/collections/asr/parts/submodules/subsampling.py | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) diff --git a/nemo/collections/asr/modules/conformer_encoder.py b/nemo/collections/asr/modules/conformer_encoder.py index 323ea3341000..6d2879e02a11 100644 --- a/nemo/collections/asr/modules/conformer_encoder.py +++ b/nemo/collections/asr/modules/conformer_encoder.py @@ -65,7 +65,7 @@ class ConformerEncoder(NeuralModule, StreamingEncoder, Exportable, AccessMixin): d_model (int): the hidden size of the model feat_out (int): the size of the output features Defaults to -1 (means feat_out is d_model) - subsampling (str): the method of subsampling, choices=['vggnet', 'striding'] + subsampling (str): the method of subsampling, choices=['vggnet', 'striding', 'dw-striding', 'stacking', 'stacking_norm'] Defaults to striding. subsampling_factor (int): the subsampling factor which should be power of 2 Defaults to 4. diff --git a/nemo/collections/asr/parts/submodules/subsampling.py b/nemo/collections/asr/parts/submodules/subsampling.py index 23bd625108c7..9e9d6726f5c0 100644 --- a/nemo/collections/asr/parts/submodules/subsampling.py +++ b/nemo/collections/asr/parts/submodules/subsampling.py @@ -64,7 +64,7 @@ class ConvSubsampling(torch.nn.Module): VGGNet Subsampling: Transformer-transducer: end-to-end speech recognition with self-attention (https://arxiv.org/pdf/1910.12977.pdf) Striding Subsampling: "Speech-Transformer: A No-Recurrence Sequence-to-Sequence Model for Speech Recognition" by Linhao Dong et al. (https://ieeexplore.ieee.org/document/8462506) Args: - subsampling (str): The subsampling technique from {"vggnet", "striding"} + subsampling (str): The subsampling technique from {"vggnet", "striding", "dw-striding"} subsampling_factor (int): The subsampling factor which should be a power of 2 subsampling_conv_chunking_factor (int): Input chunking factor which can be -1 (no chunking) 1 (auto) or a power of 2. Default is 1