Correct prompt padding side (#1)

- Correct prompt padding side (037776cca036c2b340673b03e3f25470c913938e)
- Update app.py (627dc63ff0fb3ceb1448818235c9f532da50a2b7)


Co-authored-by: Yoach Lacombe <ylacombe@users.noreply.huggingface.co>

app.py

@@ -29,7 +29,8 @@ model = ParlerTTSForConditionalGeneration.from_pretrained(
 
 client = InferenceClient()
 
-tokenizer = AutoTokenizer.from_pretrained(repo_id)
+description_tokenizer = AutoTokenizer.from_pretrained(repo_id) 
+prompt_tokenizer = AutoTokenizer.from_pretrained(repo_id, padding_side="left")
 feature_extractor = AutoFeatureExtractor.from_pretrained(repo_id)
 
 SAMPLE_RATE = feature_extractor.sampling_rate
@@ -87,8 +88,8 @@ def generate_base(subject, setting):
 
     gr.Info("Generating Audio")
     description = "Jenny speaks at an average pace with a calm delivery in a very confined sounding environment with clear audio quality."
-    story_tokens = tokenizer(model_input_tokens, return_tensors="pt", padding=True).input_ids.to(device)
-    description_tokens = tokenizer([description for _ in range(len(model_input_tokens))], return_tensors="pt").input_ids.to(device)
+    story_tokens = prompt_tokenizer(model_input_tokens, return_tensors="pt", padding=True).input_ids.to(device)
+    description_tokens = description_tokenizer([description for _ in range(len(model_input_tokens))], return_tensors="pt").input_ids.to(device)
     speech_output = model.generate(input_ids=description_tokens, prompt_input_ids=story_tokens)
     speech_output = [output.cpu().numpy() for output in speech_output]
     gr.Info("Generated Audio")