make style

init_dummy_model.py

@@ -13,7 +13,7 @@ encodec_vocab_size = encodec.codebook_size
 
 
 decoder_config = ParlerTTSDecoderConfig(
-    vocab_size=encodec_vocab_size+1,
+    vocab_size=encodec_vocab_size + 1,
     max_position_embeddings=2048,
     num_hidden_layers=4,
     ffn_dim=512,
@@ -27,34 +27,32 @@ decoder_config = ParlerTTSDecoderConfig(
     activation_dropout=0.0,
     pad_token_id=encodec_vocab_size,
     eos_token_id=encodec_vocab_size,
-    bos_token_id=encodec_vocab_size+1,
+    bos_token_id=encodec_vocab_size + 1,
     num_codebooks=num_codebooks,
 )
 # TODO: ?? how to make it stop ?
-        
 
-        
+
 decoder = ParlerTTSForCausalLM(decoder_config)
 
 decoder.save_pretrained("/raid/yoach/tmp/artefacts/decoder/")
 
 
-
 model = ParlerTTSForConditionalGeneration.from_sub_models_pretrained(
     text_encoder_pretrained_model_name_or_path=text_model,
     audio_encoder_pretrained_model_name_or_path=encodec_version,
     decoder_pretrained_model_name_or_path="/raid/yoach/tmp/artefacts/decoder/",
-    vocab_size = t5.vocab_size
+    vocab_size=t5.vocab_size,
 )
 
 # set the appropriate bos/pad token ids
-model.generation_config.decoder_start_token_id = encodec_vocab_size+1
+model.generation_config.decoder_start_token_id = encodec_vocab_size + 1
 model.generation_config.pad_token_id = encodec_vocab_size
 model.generation_config.eos_token_id = encodec_vocab_size
 
 # set other default generation config params
 model.generation_config.max_length = int(30 * model.audio_encoder.config.frame_rate)
-model.generation_config.do_sample = False # True
-model.generation_config.guidance_scale = 1 # 3.0
+model.generation_config.do_sample = False  # True
+model.generation_config.guidance_scale = 1  # 3.0
 
-model.save_pretrained("/raid/yoach/tmp/artefacts/tiny-model/")
\ No newline at end of file
+model.save_pretrained("/raid/yoach/tmp/artefacts/tiny-model/")

init_dummy_model_dac.py

@@ -20,7 +20,7 @@ encodec_vocab_size = encodec.codebook_size
 
 
 decoder_config = ParlerTTSDecoderConfig(
-    vocab_size=encodec_vocab_size+1,
+    vocab_size=encodec_vocab_size + 1,
     max_position_embeddings=2048,
     num_hidden_layers=4,
     ffn_dim=512,
@@ -34,34 +34,32 @@ decoder_config = ParlerTTSDecoderConfig(
     activation_dropout=0.0,
     pad_token_id=encodec_vocab_size,
     eos_token_id=encodec_vocab_size,
-    bos_token_id=encodec_vocab_size+1,
+    bos_token_id=encodec_vocab_size + 1,
     num_codebooks=num_codebooks,
 )
 # TODO: ?? how to make it stop ?
-        
 
-        
+
 decoder = ParlerTTSForCausalLM(decoder_config)
 
 decoder.save_pretrained("/raid/yoach/tmp/artefacts/decoder/")
 
 
-
 model = ParlerTTSForConditionalGeneration.from_sub_models_pretrained(
     text_encoder_pretrained_model_name_or_path=text_model,
     audio_encoder_pretrained_model_name_or_path=encodec_version,
     decoder_pretrained_model_name_or_path="/raid/yoach/tmp/artefacts/decoder/",
-    vocab_size = t5.vocab_size
+    vocab_size=t5.vocab_size,
 )
 
 # set the appropriate bos/pad token ids
-model.generation_config.decoder_start_token_id = encodec_vocab_size+1
+model.generation_config.decoder_start_token_id = encodec_vocab_size + 1
 model.generation_config.pad_token_id = encodec_vocab_size
 model.generation_config.eos_token_id = encodec_vocab_size
 
 # set other default generation config params
 model.generation_config.max_length = int(30 * model.audio_encoder.config.frame_rate)
-model.generation_config.do_sample = False # True
-model.generation_config.guidance_scale = 1 # 3.0
+model.generation_config.do_sample = False  # True
+model.generation_config.guidance_scale = 1  # 3.0
 
-model.save_pretrained("/raid/yoach/tmp/artefacts/tiny-dac-model/")
\ No newline at end of file
+model.save_pretrained("/raid/yoach/tmp/artefacts/tiny-dac-model/")

init_model.py

@@ -21,8 +21,8 @@ encodec_vocab_size = encodec.codebook_size
 
 
 decoder_config = ParlerTTSDecoderConfig(
-    vocab_size=encodec_vocab_size+1,
-    max_position_embeddings=3000, # 30 s = 2580
+    vocab_size=encodec_vocab_size + 1,
+    max_position_embeddings=3000,  # 30 s = 2580
     num_hidden_layers=12,
     ffn_dim=4096,
     num_attention_heads=16,
@@ -35,33 +35,31 @@ decoder_config = ParlerTTSDecoderConfig(
     activation_dropout=0.0,
     pad_token_id=encodec_vocab_size,
     eos_token_id=encodec_vocab_size,
-    bos_token_id=encodec_vocab_size+1,
+    bos_token_id=encodec_vocab_size + 1,
     num_codebooks=num_codebooks,
 )
 
-        
+
 decoder = ParlerTTSForCausalLM(decoder_config)
 decoder.save_pretrained("/raid/yoach/tmp/artefacts/decoder/")
 
 
-
-
 model = ParlerTTSForConditionalGeneration.from_sub_models_pretrained(
     text_encoder_pretrained_model_name_or_path=text_model,
     audio_encoder_pretrained_model_name_or_path=encodec_version,
     decoder_pretrained_model_name_or_path="/raid/yoach/tmp/artefacts/decoder/",
-    vocab_size = t5.vocab_size
+    vocab_size=t5.vocab_size,
 )
 
 # set the appropriate bos/pad token ids
-model.generation_config.decoder_start_token_id = encodec_vocab_size+1
+model.generation_config.decoder_start_token_id = encodec_vocab_size + 1
 model.generation_config.pad_token_id = encodec_vocab_size
 model.generation_config.eos_token_id = encodec_vocab_size
 
 # set other default generation config params
 model.generation_config.max_length = int(30 * model.audio_encoder.config.frame_rate)
-model.generation_config.do_sample = False # True
-model.generation_config.guidance_scale = 1 # 3.0
+model.generation_config.do_sample = False  # True
+model.generation_config.guidance_scale = 1  # 3.0
 
 
-model.save_pretrained("/raid/yoach/tmp/artefacts/small-stable-speech-untrained/")
\ No newline at end of file
+model.save_pretrained("/raid/yoach/tmp/artefacts/small-stable-speech-untrained/")

init_model_75M.py

@@ -21,8 +21,8 @@ encodec_vocab_size = encodec.codebook_size
 
 
 decoder_config = ParlerTTSDecoderConfig(
-    vocab_size=encodec_vocab_size+1,
-    max_position_embeddings=4096, # 30 s = 2580
+    vocab_size=encodec_vocab_size + 1,
+    max_position_embeddings=4096,  # 30 s = 2580
     num_hidden_layers=8,
     ffn_dim=3072,
     num_attention_heads=12,
@@ -35,33 +35,31 @@ decoder_config = ParlerTTSDecoderConfig(
     activation_dropout=0.0,
     pad_token_id=encodec_vocab_size,
     eos_token_id=encodec_vocab_size,
-    bos_token_id=encodec_vocab_size+1,
+    bos_token_id=encodec_vocab_size + 1,
     num_codebooks=num_codebooks,
 )
 
-        
+
 decoder = ParlerTTSForCausalLM(decoder_config)
 decoder.save_pretrained("/raid/yoach/tmp/artefacts/decoder_small/")
 
 
-
-
 model = ParlerTTSForConditionalGeneration.from_sub_models_pretrained(
     text_encoder_pretrained_model_name_or_path=text_model,
     audio_encoder_pretrained_model_name_or_path=encodec_version,
     decoder_pretrained_model_name_or_path="/raid/yoach/tmp/artefacts/decoder_small/",
-    vocab_size = t5.vocab_size
+    vocab_size=t5.vocab_size,
 )
 
 # set the appropriate bos/pad token ids
-model.generation_config.decoder_start_token_id = encodec_vocab_size+1
+model.generation_config.decoder_start_token_id = encodec_vocab_size + 1
 model.generation_config.pad_token_id = encodec_vocab_size
 model.generation_config.eos_token_id = encodec_vocab_size
 
 # set other default generation config params
 model.generation_config.max_length = int(30 * model.audio_encoder.config.frame_rate)
-model.generation_config.do_sample = False # True
-model.generation_config.guidance_scale = 1 # 3.0
+model.generation_config.do_sample = False  # True
+model.generation_config.guidance_scale = 1  # 3.0
 
 
-model.save_pretrained("/raid/yoach/tmp/artefacts/stable-speech-untrained-75M/")
\ No newline at end of file
+model.save_pretrained("/raid/yoach/tmp/artefacts/stable-speech-untrained-75M/")

parler_tts/__init__.py

 from .configuration_parler_tts import ParlerTTSConfig, ParlerTTSDecoderConfig
-from .modeling_parler_tts import ParlerTTSForCausalLM, ParlerTTSForConditionalGeneration, apply_delay_pattern_mask, build_delay_pattern_mask
+from .modeling_parler_tts import (
+    ParlerTTSForCausalLM,
+    ParlerTTSForConditionalGeneration,
+    apply_delay_pattern_mask,
+    build_delay_pattern_mask,
+)
 
-from .dac_wrapper import DACConfig, DACModel
\ No newline at end of file
+from .dac_wrapper import DACConfig, DACModel

parler_tts/configuration_parler_tts.py

@@ -81,7 +81,7 @@ class ParlerTTSDecoderConfig(PretrainedConfig):
 
     def __init__(
         self,
-        vocab_size=2049, # vocab size = 2048 (encodec vocab size) + 1 (eos)
+        vocab_size=2049,  # vocab size = 2048 (encodec vocab size) + 1 (eos)
         max_position_embeddings=2048,
         num_hidden_layers=24,
         ffn_dim=4096,

parler_tts/dac_wrapper/__init__.py

 from .configuration_dac import DACConfig
-from .modeling_dac import DACModel
\ No newline at end of file
+from .modeling_dac import DACModel

parler_tts/dac_wrapper/configuration_dac.py

@@ -8,17 +8,16 @@ class DACConfig(PretrainedConfig):
     def __init__(
         self,
         num_codebooks: int = 9,
-        model_bitrate: int = 8, # kbps
+        model_bitrate: int = 8,  # kbps
         codebook_size: int = 1024,
         latent_dim: int = 1024,
         frame_rate: int = 86,
         **kwargs,
     ):
-
         self.codebook_size = codebook_size
         self.model_bitrate = model_bitrate
         self.latent_dim = latent_dim
         self.num_codebooks = num_codebooks
         self.frame_rate = frame_rate
-        
-        super().__init__(**kwargs)
\ No newline at end of file
+
+        super().__init__(**kwargs)

parler_tts/dac_wrapper/modeling_dac.py

@@ -7,22 +7,24 @@ from .configuration_dac import DACConfig
 from dac.model import DAC
 
 
+# model doesn't support batching yet
 
-# model doesn't support batching yet 
 
 class DACModel(PreTrainedModel):
     config_class = DACConfig
 
     def __init__(self, config):
         super().__init__(config)
-        
+
         self.model = DAC(
-            n_codebooks = config.num_codebooks,
-            latent_dim = config.latent_dim,
-            codebook_size = config.codebook_size,
+            n_codebooks=config.num_codebooks,
+            latent_dim=config.latent_dim,
+            codebook_size=config.codebook_size,
         )
-        
-    def encode(self, input_values, padding_mask=None, bandwidth=None, return_dict=None, n_quantizers=None, sample_rate=None):
+
+    def encode(
+        self, input_values, padding_mask=None, bandwidth=None, return_dict=None, n_quantizers=None, sample_rate=None
+    ):
         """
         Encodes the input audio waveform into discrete codes.
 
@@ -44,7 +46,7 @@ class DACModel(PreTrainedModel):
             factors for each chunk when `normalize` is True. Each frames is a tuple `(codebook, scale)`, with
             `codebook` of shape `[batch_size, num_codebooks, frames]`.
             Scale is not used here.
-    
+
         """
         _, channels, input_length = input_values.shape
 
@@ -52,12 +54,12 @@ class DACModel(PreTrainedModel):
             raise ValueError(f"Number of audio channels must be 1 or 2, but got {channels}")
 
         audio_data = self.model.preprocess(input_values, sample_rate)
-        
+
         return_dict = return_dict if return_dict is not None else self.config.return_dict
 
         # TODO: for now, no chunk length
 
-        chunk_length = None # self.config.chunk_length
+        chunk_length = None  # self.config.chunk_length
         if chunk_length is None:
             chunk_length = input_length
             stride = input_length
@@ -79,9 +81,9 @@ class DACModel(PreTrainedModel):
         for offset in range(0, input_length - step, stride):
             mask = padding_mask[..., offset : offset + chunk_length].bool()
             frame = audio_data[:, :, offset : offset + chunk_length]
-            
+
             scale = None
-            
+
             _, encoded_frame, _, _, _ = self.model.encode(frame, n_quantizers=n_quantizers)
             encoded_frames.append(encoded_frame)
             scales.append(scale)
@@ -92,15 +94,14 @@ class DACModel(PreTrainedModel):
             return (encoded_frames, scales)
 
         return EncodecEncoderOutput(encoded_frames, scales)
-    
-    
+
     def decode(
-            self,
-            audio_codes,
-            audio_scales,
-            padding_mask = None,
-            return_dict = None,
-        ):
+        self,
+        audio_codes,
+        audio_scales,
+        padding_mask=None,
+        return_dict=None,
+    ):
         """
         Decodes the given frames into an output audio waveform.
 
@@ -125,12 +126,12 @@ class DACModel(PreTrainedModel):
 
         if len(audio_codes) != 1:
             raise ValueError(f"Expected one frame, got {len(audio_codes)}")
-        
+
         audio_values = self.model.quantizer.from_codes(audio_codes.squeeze(0))[0]
         audio_values = self.model.decode(audio_values)
         if not return_dict:
             return (audio_values,)
         return EncodecDecoderOutput(audio_values)
-    
+
     def forward(self, tensor):
-        raise ValueError(f"`DACModel.forward` not implemented yet")
\ No newline at end of file
+        raise ValueError(f"`DACModel.forward` not implemented yet")

push_dac_to_hub.py

-
 import dac
+
 # Download a model
 model_path = dac.utils.download(model_type="44khz")
 model = dac.DAC.load(model_path)
@@ -10,6 +10,7 @@ hf_dac = DACModel(DACConfig())
 hf_dac.model.load_state_dict(model.state_dict())
 
 from transformers import AutoConfig, AutoModel
+
 AutoConfig.register("dac", DACConfig)
 AutoModel.register(DACConfig, DACModel)
 
@@ -20,4 +21,4 @@ hf_dac.push_to_hub("ylacombe/dac_44khZ_8kbps")
 
 from transformers import EncodecFeatureExtractor
 
-EncodecFeatureExtractor(sampling_rate=44100).push_to_hub("ylacombe/dac_44khZ_8kbps")
\ No newline at end of file
+EncodecFeatureExtractor(sampling_rate=44100).push_to_hub("ylacombe/dac_44khZ_8kbps")