API request failed with status code 502

[TabNahida]

🖥️ Hardware: CPU only (No CUDA GPU detected)
📊 CPU: 16 cores, 32 threads
📊 RAM: 189.55 GB
⚙️ Using CPU-optimized settings
Configuration loaded:
  API_URL: http://localhost:1234/v1/completions
  MAX_TOKENS: 8192
  TEMPERATURE: 0.6
  TOP_P: 0.9
  REPETITION_PENALTY: 1.1
🔥 Starting Orpheus-FASTAPI Server on 0.0.0.0:5005
💬 Web UI available at http://localhost:5005
📖 API docs available at http://localhost:5005/docs
🔗 Using LLM inference server at: http://localhost:1234/v1/completions
INFO:     Will watch for changes in these directories: ['C:\\Project\\AI\\Orpheus-FastAPI']
INFO:     Uvicorn running on http://0.0.0.0:5005 (Press CTRL+C to quit)
INFO:     Started reloader process [30352] using WatchFiles
INFO:     Started server process [43848]
INFO:     Waiting for application startup.
INFO:     Application startup complete.
INFO:     127.0.0.1:64617 - "GET / HTTP/1.1" 200 OK
INFO:     127.0.0.1:64617 - "GET /get_config HTTP/1.1" 200 OK
Starting speech generation for 'A'
Using voice: tara, GPU acceleration: No
Generating speech for: <|audio|>tara: A<|eot_id|>
Error: API request failed with status code 502
Error details:
Retrying in 2 seconds...
Error: API request failed with status code 502
Error details:
Retrying in 4 seconds...
Error: API request failed with status code 502
Error details:
Retrying in 8 seconds...
Producer completed - setting done event
Received end-of-stream marker
Waiting for token processor thread to complete...
Audio saved to outputs/tara_20250825_052658.wav
Total speech generation completed in 14.59 seconds
INFO:     127.0.0.1:64618 - "POST /speak HTTP/1.1" 200 OK
INFO:     127.0.0.1:64618 - "GET /tara_20250825_052658.wav HTTP/1.1" 200 OK

I'm using llama.cpp as my server.

./llama-server -m D:\Data\AI\TTS\Models-GGUF\Orpheus-3b-FT-Q8_0.gguf --ctx-size 8192 --n-predict 8192 --rope-scaling linear --port 1234

I don't why. I use the same parameters as README