Spaces:
Running
Running
updated the app.py
Browse files
app.py
CHANGED
|
@@ -5,7 +5,8 @@ import logging
|
|
| 5 |
from fastapi import FastAPI, HTTPException, status
|
| 6 |
from pydantic import BaseModel
|
| 7 |
from nemo.collections.tts.models import FastPitchModel, HifiGanModel
|
| 8 |
-
|
|
|
|
| 9 |
|
| 10 |
# Configure logging
|
| 11 |
logging.basicConfig(level=logging.INFO)
|
|
@@ -39,65 +40,22 @@ def load_models():
|
|
| 39 |
models['en'].eval()
|
| 40 |
logger.info("English model loaded successfully")
|
| 41 |
|
| 42 |
-
# Load the Bikol Spectrogram Generator
|
| 43 |
logger.info("Loading Bikol FastPitch model...")
|
| 44 |
-
|
| 45 |
-
|
| 46 |
-
|
| 47 |
-
|
| 48 |
-
|
| 49 |
-
).to(device)
|
| 50 |
-
models['bikol'].eval()
|
| 51 |
-
logger.info("Bikol model loaded successfully")
|
| 52 |
-
|
| 53 |
-
except Exception as e:
|
| 54 |
-
logger.warning(f"First attempt failed: {e}")
|
| 55 |
-
logger.info("Attempting to load Bikol model with config override...")
|
| 56 |
-
|
| 57 |
-
# Second attempt: Override the text_tokenizer config to remove g2p parameter
|
| 58 |
-
try:
|
| 59 |
-
# Create a config override that removes the problematic g2p parameter
|
| 60 |
-
override_config = OmegaConf.create({
|
| 61 |
-
'text_tokenizer': {
|
| 62 |
-
'_target_': 'nemo.collections.common.tokenizers.text_to_speech.tts_tokenizers.BaseCharsTokenizer',
|
| 63 |
-
'pad_with_space': True
|
| 64 |
-
}
|
| 65 |
-
})
|
| 66 |
-
|
| 67 |
-
models['bikol'] = FastPitchModel.restore_from(
|
| 68 |
-
"models/fastpitch_bikol_fixed.nemo",
|
| 69 |
-
override_config_path=override_config,
|
| 70 |
-
strict=False
|
| 71 |
-
).to(device)
|
| 72 |
-
models['bikol'].eval()
|
| 73 |
-
logger.info("Bikol model loaded successfully with config override")
|
| 74 |
-
|
| 75 |
-
except Exception as e2:
|
| 76 |
-
logger.error(f"Failed to load Bikol model with override: {e2}")
|
| 77 |
-
# Third attempt: Try modifying the saved config
|
| 78 |
-
logger.info("Attempting alternative loading method...")
|
| 79 |
-
|
| 80 |
-
try:
|
| 81 |
-
# Load model with map_location to avoid device issues
|
| 82 |
-
models['bikol'] = FastPitchModel.restore_from(
|
| 83 |
-
"models/fastpitch_bikol_fixed.nemo",
|
| 84 |
-
map_location=device,
|
| 85 |
-
strict=False
|
| 86 |
-
)
|
| 87 |
-
models['bikol'].eval()
|
| 88 |
-
logger.info("Bikol model loaded with map_location")
|
| 89 |
-
except Exception as e3:
|
| 90 |
-
logger.error(f"All attempts to load Bikol model failed: {e3}")
|
| 91 |
-
logger.error("Bikol language will not be available")
|
| 92 |
-
# Don't raise - allow app to start with just English
|
| 93 |
-
|
| 94 |
-
logger.info("Model loading complete. Available models: " + ", ".join(models.keys()))
|
| 95 |
-
|
| 96 |
except Exception as e:
|
| 97 |
logger.error(f"FATAL: Could not load models. Error: {e}")
|
| 98 |
import traceback
|
| 99 |
traceback.print_exc()
|
| 100 |
-
#
|
|
|
|
|
|
|
|
|
|
|
|
|
| 101 |
|
| 102 |
# --- 3. Define API Request and Response Models ---
|
| 103 |
class TTSRequest(BaseModel):
|
|
@@ -110,7 +68,7 @@ def synthesize_speech(request: TTSRequest):
|
|
| 110 |
"""
|
| 111 |
Generates speech from text using the selected language model.
|
| 112 |
"""
|
| 113 |
-
if not models:
|
| 114 |
raise HTTPException(
|
| 115 |
status_code=status.HTTP_503_SERVICE_UNAVAILABLE,
|
| 116 |
detail="Models are not loaded yet. Please try again in a moment."
|
|
|
|
| 5 |
from fastapi import FastAPI, HTTPException, status
|
| 6 |
from pydantic import BaseModel
|
| 7 |
from nemo.collections.tts.models import FastPitchModel, HifiGanModel
|
| 8 |
+
# Omegaconf is no longer needed here since we aren't creating overrides
|
| 9 |
+
# from omegaconf import OmegaConf, open_dict
|
| 10 |
|
| 11 |
# Configure logging
|
| 12 |
logging.basicConfig(level=logging.INFO)
|
|
|
|
| 40 |
models['en'].eval()
|
| 41 |
logger.info("English model loaded successfully")
|
| 42 |
|
| 43 |
+
# Load the CORRECTED Bikol Spectrogram Generator
|
| 44 |
logger.info("Loading Bikol FastPitch model...")
|
| 45 |
+
# This is the only line needed now. Replace the filename with your new .nemo file.
|
| 46 |
+
models['bikol'] = FastPitchModel.restore_from("models/fastpitch_bikol_corrected.nemo").to(device)
|
| 47 |
+
models['bikol'].eval()
|
| 48 |
+
logger.info("Bikol model loaded successfully")
|
| 49 |
+
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 50 |
except Exception as e:
|
| 51 |
logger.error(f"FATAL: Could not load models. Error: {e}")
|
| 52 |
import traceback
|
| 53 |
traceback.print_exc()
|
| 54 |
+
# You might want the app to fail completely if models don't load
|
| 55 |
+
# raise e
|
| 56 |
+
|
| 57 |
+
logger.info("Model loading complete. Available models: " + ", ".join(models.keys()))
|
| 58 |
+
|
| 59 |
|
| 60 |
# --- 3. Define API Request and Response Models ---
|
| 61 |
class TTSRequest(BaseModel):
|
|
|
|
| 68 |
"""
|
| 69 |
Generates speech from text using the selected language model.
|
| 70 |
"""
|
| 71 |
+
if not models or 'hifigan' not in models:
|
| 72 |
raise HTTPException(
|
| 73 |
status_code=status.HTTP_503_SERVICE_UNAVAILABLE,
|
| 74 |
detail="Models are not loaded yet. Please try again in a moment."
|