- Added JsonProcessor module to handle Python dictionary to JSON conversion
- Enhanced response processing to automatically format single quotes and
None values
- Updated requirements.txt to include loguru dependency
- Model responses now support both Python dict and JSON formats seamlessly
perf: use safetensors for faster model loading.
yongdong
9d3765e
fix: add explicit temperature and top_p parameters to avoid warnings.
yongdong
51982fa
Update examples.
yongdong
0b315c1
Update lora model name.
yongdong
6158c62
Disable sampling for deterministic JSON output
yongdong
1f6f70b
Update Readme.md
yongdong
c6b828a
feat: enforce JSON-only output in generate_response_gpu prompt