fix comment, test=doc

speechx/speechx/decoder/ctc_prefix_beam_search_decoder_main.cc

@@ -30,10 +30,10 @@ DEFINE_string(dict_file, "vocab.txt", "vocabulary of lm");
 DEFINE_string(lm_path, "", "language model");
 DEFINE_int32(receptive_field_length,
              7,
-             "receptive field of two CNN(kernel=5) downsampling module.");
+             "receptive field of two CNN(kernel=3) downsampling module.");
 DEFINE_int32(downsampling_rate,
              4,
-             "two CNN(kernel=5) module downsampling rate.");
+             "two CNN(kernel=3) module downsampling rate.");
 DEFINE_string(
     model_input_names,
     "audio_chunk,audio_chunk_lens,chunk_state_h_box,chunk_state_c_box",

speechx/speechx/decoder/param.h

@@ -28,10 +28,10 @@ DEFINE_string(cmvn_file, "", "read cmvn");
 // feature sliding window
 DEFINE_int32(receptive_field_length,
              7,
-             "receptive field of two CNN(kernel=5) downsampling module.");
+             "receptive field of two CNN(kernel=3) downsampling module.");
 DEFINE_int32(downsampling_rate,
              4,
-             "two CNN(kernel=5) module downsampling rate.");
+             "two CNN(kernel=3) module downsampling rate.");
 DEFINE_int32(nnet_decoder_chunk, 1, "paddle nnet forward chunk");
 // nnet
 DEFINE_string(model_path, "avg_1.jit.pdmodel", "paddle nnet model");

speechx/speechx/decoder/tlg_decoder_main.cc

@@ -33,10 +33,10 @@ DEFINE_int32(max_active, 7500, "decoder graph");
 DEFINE_int32(nnet_decoder_chunk, 1, "paddle nnet forward chunk");
 DEFINE_int32(receptive_field_length,
              7,
-             "receptive field of two CNN(kernel=5) downsampling module.");
+             "receptive field of two CNN(kernel=3) downsampling module.");
 DEFINE_int32(downsampling_rate,
              4,
-             "two CNN(kernel=5) module downsampling rate.");
+             "two CNN(kernel=3) module downsampling rate.");
 DEFINE_string(
     model_input_names,
     "audio_chunk,audio_chunk_lens,chunk_state_h_box,chunk_state_c_box",

speechx/speechx/nnet/nnet_forward_main.cc

@@ -27,10 +27,10 @@ DEFINE_string(param_path, "avg_1.jit.pdiparams", "paddle nnet model param");
 DEFINE_int32(nnet_decoder_chunk, 1, "paddle nnet forward chunk");
 DEFINE_int32(receptive_field_length,
              7,
-             "receptive field of two CNN(kernel=5) downsampling module.");
+             "receptive field of two CNN(kernel=3) downsampling module.");
 DEFINE_int32(downsampling_rate,
              4,
-             "two CNN(kernel=5) module downsampling rate.");
+             "two CNN(kernel=3) module downsampling rate.");
 DEFINE_string(
     model_input_names,
     "audio_chunk,audio_chunk_lens,chunk_state_h_box,chunk_state_c_box",
@@ -159,4 +159,4 @@ int main(int argc, char* argv[]) {
     KALDI_LOG << "Done " << num_done << " utterances, " << num_err
               << " with errors.";
     return (num_done != 0 ? 0 : 1);
-}
+}