File size: 5,386 Bytes
22d5f88
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1b96d91
22d5f88
1b96d91
22d5f88
 
 
 
 
1b96d91
 
 
 
22d5f88
 
1b96d91
 
 
 
 
22d5f88
 
 
 
 
 
 
 
1b96d91
22d5f88
 
 
 
 
 
 
 
 
 
 
 
 
 
 
44aa5c1
22d5f88
 
 
 
 
 
 
 
581f5e4
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
22d5f88
 
04cd4ef
 
 
 
 
 
 
22d5f88
04cd4ef
 
 
 
 
22d5f88
04cd4ef
22d5f88
04cd4ef
 
 
 
22d5f88
 
04cd4ef
22d5f88
04cd4ef
22d5f88
 
 
 
 
 
 
 
04cd4ef
 
 
 
 
 
 
343c6ad
22d5f88
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
from fastapi import FastAPI, HTTPException
import numpy as np
import torch
from pydantic import BaseModel
import base64
import io
import os
import logging
from pathlib import Path
from inference import InferenceRecipe
from fastapi.middleware.cors import CORSMiddleware

logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)

app = FastAPI()

# Add CORS middleware
app.add_middleware(
    CORSMiddleware,
    allow_origins=["*"],
    allow_credentials=True,
    allow_methods=["*"],
    allow_headers=["*"],
)

class AudioRequest(BaseModel):
    audio_data: str
    sample_rate: int

class AudioResponse(BaseModel):
    audio_data: str
    text: str = ""

# Model initialization status
INITIALIZATION_STATUS = {
    "model_loaded": False,
    "error": None
}

# Global model instance
model = None


def initialize_model():
    """Initialize the model with correct path resolution"""
    global model, INITIALIZATION_STATUS
    try:
        device = "cuda" if torch.cuda.is_available() else "cpu"
        logger.info(f"Initializing model on device: {device}")
        
        # Critical: Use absolute path for model loading
        model_path = os.path.abspath(os.path.join('/app/src', 'models'))
        logger.info(f"Loading models from: {model_path}")
        
        if not os.path.exists(model_path):
            raise RuntimeError(f"Model path {model_path} does not exist")
            
        # Log available model files for debugging
        model_files = os.listdir(model_path)
        logger.info(f"Available model files: {model_files}")
        
        model = InferenceRecipe(model_path, device=device)
        INITIALIZATION_STATUS["model_loaded"] = True
        logger.info("Model initialized successfully")
        return True
    except Exception as e:
        INITIALIZATION_STATUS["error"] = str(e)
        logger.error(f"Failed to initialize model: {e}")
        return False
        
@app.on_event("startup")
async def startup_event():
    """Initialize model on startup"""
    initialize_model()

@app.get("/api/v1/health")
def health_check():
    """Health check endpoint"""
    status = {
        "status": "healthy" if INITIALIZATION_STATUS["model_loaded"] else "initializing",
        "initialization_status": INITIALIZATION_STATUS
    }
    
    if model is not None:
        status.update({
            "device": "cpu",
            "model_path": str(model.model_path),
            "mimi_loaded": model.mimi is not None,
            "tokenizer_loaded": model.text_tokenizer is not None,
            "lm_loaded": model.lm_gen is not None
        })
        
    return status

# @app.post("/api/v1/inference")
# async def inference(request: AudioRequest) -> AudioResponse:
#     """Run inference on audio input"""
#     if not INITIALIZATION_STATUS["model_loaded"]:
#         raise HTTPException(
#             status_code=503,
#             detail=f"Model not ready. Status: {INITIALIZATION_STATUS}"
#         )
        
#     try:
#         # Decode audio from base64
#         audio_bytes = base64.b64decode(request.audio_data) 
#         audio_array = np.load(io.BytesIO(audio_bytes))
        
#         # Run inference
#         result = model.inference(audio_array, request.sample_rate)
        
#         # Encode output audio
#         buffer = io.BytesIO()
#         np.save(buffer, result['audio'])
#         audio_b64 = base64.b64encode(buffer.getvalue()).decode()
        
#         return AudioResponse(
#             audio_data=audio_b64,
#             text=result.get("text", "")
#         )
#     except Exception as e:
#         logger.error(f"Inference failed: {str(e)}")
#         raise HTTPException(status_code=500, detail=str(e))
@app.post("/api/v1/inference")
async def inference(request: AudioRequest) -> AudioResponse:
    """Run inference with enhanced error handling and logging"""
    if not INITIALIZATION_STATUS["model_loaded"]:
        raise HTTPException(
            status_code=503,
            detail=f"Model not ready. Status: {INITIALIZATION_STATUS}"
        )
        
    try:
        # Log input validation
        logger.info(f"Received inference request with sample rate: {request.sample_rate}")
        
        # Decode audio
        audio_bytes = base64.b64decode(request.audio_data)
        audio_array = np.load(io.BytesIO(audio_bytes))
        logger.info(f"Decoded audio array shape: {audio_array.shape}, dtype: {audio_array.dtype}")
        
        # Validate input format
        if len(audio_array.shape) != 2:
            raise ValueError(f"Expected 2D audio array [C,T], got shape {audio_array.shape}")
            
        # Run inference
        result = model.inference(audio_array, request.sample_rate)
        logger.info(f"Inference complete. Output shape: {result['audio'].shape}")
        
        # Encode output
        buffer = io.BytesIO()
        np.save(buffer, result['audio'])
        audio_b64 = base64.b64encode(buffer.getvalue()).decode()
        
        return AudioResponse(
            audio_data=audio_b64,
            text=result.get("text", "")
        )
        
    except Exception as e:
        logger.error(f"Inference failed: {str(e)}", exc_info=True)
        raise HTTPException(
            status_code=500, 
            detail=str(e)
        )
        
if __name__ == "__main__":
    import uvicorn
    uvicorn.run(app, host="0.0.0.0", port=8000)