Add inputs_to_logits_ratio to LasrCTCConfig

src/transformers/models/lasr/configuration_lasr.py

@@ -185,7 +185,7 @@ class LasrCTCConfig(PreTrainedConfig):
             encoder_config (`Union[dict, LasrEncoderConfig]`, *optional*):
                 The config object or dictionary of the encoder.
             pad_token_id (`int`, *optional*, defaults to 0):
-                Padding token id. Also used as blank token id.
+                Padding token id. Also used as blank token id..
     Example:
         ```python
         >>> from transformers import LasrForCTC, LasrCTCConfig
@@ -240,5 +240,9 @@ def from_encoder_config(cls, encoder_config: LasrEncoderConfig, **kwargs):
 
         return cls(encoder_config=encoder_config.to_dict(), **kwargs)
 
+    @property
+    def inputs_to_logits_ratio(self):
+        return self.encoder_config.subsampling_conv_stride**2
+
 
 __all__ = ["LasrEncoderConfig", "LasrCTCConfig"]

src/transformers/models/lasr/modular_lasr.py

@@ -258,7 +258,7 @@ class LasrCTCConfig(ParakeetCTCConfig):
             encoder_config (`Union[dict, LasrEncoderConfig]`, *optional*):
                 The config object or dictionary of the encoder.
             pad_token_id (`int`, *optional*, defaults to 0):
-                Padding token id. Also used as blank token id.
+                Padding token id. Also used as blank token id..
     Example:
         ```python
         >>> from transformers import LasrForCTC, LasrCTCConfig
@@ -291,6 +291,10 @@ def __init__(
             **kwargs,
         )
 
+    @property
+    def inputs_to_logits_ratio(self):
+        return self.encoder_config.subsampling_conv_stride**2
+
 
 class LasrEncoderSubsampling(nn.Module):
     def __init__(self, config: LasrEncoderConfig):

src/transformers/pipelines/automatic_speech_recognition.py

@@ -196,6 +196,8 @@ def __init__(
         # set the model type so we can check we have the right pre- and post-processing parameters
         if model.config.model_type == "whisper":
             self.type = "seq2seq_whisper"
+        elif model.config.model_type == "lasr_ctc":
+            self.type = "lasr_ctc"
         elif model.__class__.__name__ in MODEL_FOR_SPEECH_SEQ_2_SEQ_MAPPING_NAMES.values():
             self.type = "seq2seq"
         elif (
@@ -448,9 +450,16 @@ def preprocess(self, inputs, chunk_length_s=0, stride_length_s=None):
             # Currently chunking is not possible at this level for `seq2seq` so
             # it's ok.
             align_to = getattr(self.model.config, "inputs_to_logits_ratio", 1)
-            chunk_len = int(round(chunk_length_s * self.feature_extractor.sampling_rate / align_to) * align_to)
-            stride_left = int(round(stride_length_s[0] * self.feature_extractor.sampling_rate / align_to) * align_to)
-            stride_right = int(round(stride_length_s[1] * self.feature_extractor.sampling_rate / align_to) * align_to)
+
+            if self.type == "lasr_ctc":
+                # TODO: find a standard for that but not easy because input length -> mel length depends on the feature extractor
+                # specific way of doing it
+                # means the model take mel features as input, we align according to the hop length
+                align_to *= self.feature_extractor.hop_length
+
+            chunk_len = int(round(chunk_length_s * self.feature_extractor.sampling_rate / align_to))
+            stride_left = int(round(stride_length_s[0] * self.feature_extractor.sampling_rate / align_to))
+            stride_right = int(round(stride_length_s[1] * self.feature_extractor.sampling_rate / align_to))
 
             if chunk_len < stride_left + stride_right:
                 raise ValueError("Chunk length must be superior to stride length")