Pijush2023 commited on
Commit
4de005f
·
verified ·
1 Parent(s): 6d5ef4a

Update app.py

Browse files
Files changed (1) hide show
  1. app.py +10 -9
app.py CHANGED
@@ -630,11 +630,11 @@ from langchain.chains import RetrievalQA
630
  from langchain.chains.conversation.memory import ConversationBufferWindowMemory
631
  from langchain.agents import Tool, initialize_agent
632
  from huggingface_hub import login
633
- from parler_tts import ParlerTTSForConditionalGeneration
634
- from transformers import AutoTokenizer, AutoFeatureExtractor, set_seed
635
 
636
  from transformers.models.speecht5.number_normalizer import EnglishNumberNormalizer
637
  from string import punctuation
 
 
638
 
639
  device = "cuda:0" if torch.cuda.is_available() else "cpu"
640
 
@@ -672,7 +672,9 @@ def generate_audio(text, description="Thomas speaks with emphasis and excitement
672
  prompt = tokenizer(preprocess(text), return_tensors="pt").to(device)
673
 
674
  set_seed(SEED)
675
- generation = model.generate(input_ids=inputs.input_ids, prompt_input_ids=prompt.input_ids)
 
 
676
  audio_arr = generation.cpu().numpy().squeeze()
677
 
678
  with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as f:
@@ -682,6 +684,11 @@ def generate_audio(text, description="Thomas speaks with emphasis and excitement
682
  logging.debug(f"Audio saved to {temp_audio_path}")
683
  return temp_audio_path
684
 
 
 
 
 
 
685
 
686
  # Check if the token is already set in the environment variables
687
  hf_token = os.getenv("HF_TOKEN")
@@ -1223,9 +1230,3 @@ with gr.Blocks(theme='Pijush2023/scikit-learn-pijush') as demo:
1223
 
1224
  demo.queue()
1225
  demo.launch(share=True)
1226
-
1227
-
1228
-
1229
-
1230
-
1231
-
 
630
  from langchain.chains.conversation.memory import ConversationBufferWindowMemory
631
  from langchain.agents import Tool, initialize_agent
632
  from huggingface_hub import login
 
 
633
 
634
  from transformers.models.speecht5.number_normalizer import EnglishNumberNormalizer
635
  from string import punctuation
636
+ from parler_tts import ParlerTTSForConditionalGeneration
637
+ from transformers import AutoTokenizer, AutoFeatureExtractor, set_seed
638
 
639
  device = "cuda:0" if torch.cuda.is_available() else "cpu"
640
 
 
672
  prompt = tokenizer(preprocess(text), return_tensors="pt").to(device)
673
 
674
  set_seed(SEED)
675
+ input_features = model.get_input_features(prompt.input_ids) # Ensure we have input_features
676
+
677
+ generation = model.generate(input_features=input_features, input_ids=inputs.input_ids)
678
  audio_arr = generation.cpu().numpy().squeeze()
679
 
680
  with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as f:
 
684
  logging.debug(f"Audio saved to {temp_audio_path}")
685
  return temp_audio_path
686
 
687
+ def install_parler_tts():
688
+ subprocess.check_call([sys.executable, "-m", "pip", "install", "git+https://github.com/huggingface/parler-tts.git"])
689
+
690
+ # Call the function to install parler-tts
691
+ install_parler_tts()
692
 
693
  # Check if the token is already set in the environment variables
694
  hf_token = os.getenv("HF_TOKEN")
 
1230
 
1231
  demo.queue()
1232
  demo.launch(share=True)