modify config file to match model implementation

config.yaml

@@ -12,22 +12,18 @@ encoder_conf:
     dropout_rate: 0.1
     positional_dropout_rate: 0.1
     attention_dropout_rate: 0.1
-    input_layer: 'depthwise' # encoder input type, you can chose conv2d, conv2d6 and conv2d8
+    input_layer: 'dw_striding' # encoder input type, you can chose conv2d, conv2d6 and conv2d8
     normalize_before: true
     cnn_module_kernel: 15
     use_cnn_module: true
     activation_type: 'swish'
-    pos_enc_layer_type: 'stream_rel_pos'
-    selfattention_layer_type: 'stream_rel_selfattn'
+    pos_enc_layer_type: 'chunk_rel_pos'
+    selfattention_layer_type: 'chunk_rel_seflattn'
     causal: false
     use_dynamic_chunk: false
-    use_limited_chunk: false
-    use_context_hint_chunk: false
-    right_context_probs: [0.75]
-    right_context_sizes: [128, 128, 128]
-    limited_decoding_chunk_sizes: [64, 128, 256]
-    limited_left_chunk_sizes: [128, 256, 128]
+    dynamic_chunk_sizes: [64, 128, 256]
+    dynamic_right_context_sizes: [128, 128, 128]
+    dynamic_left_context_sizes: [128, 256, 128]
     cnn_module_norm: 'layer_norm' # using nn.LayerNorm makes model converge faster
     use_dynamic_left_chunk: false
-    use_dynamic_conv: true
-    freeze_subsampling_layer: false
+    dynamic_conv: true