CLI Reference¶
PyPTO Serving exposes command-line tools for serving, offline generation, DeepSeek V4 checkpoint preparation, and runtime integration.
Installed Commands¶
| Command | Purpose |
|---|---|
pypto-serving |
Start the OpenAI-compatible HTTP server, or run offline generation when --prompt is set. |
pypto-prepack-deepseek-v4 |
Build the optional DeepSeek V4 rank-stacked weight sidecar used by the serving loader. |
Repository Utilities¶
| Command | Purpose |
|---|---|
python scripts/convert_deepseek_v4_to_w8a8.py |
Convert the DeepSeek V4 Flash source checkpoint variant validated by this repository to the W8A8 layout expected by PyPTO Serving. |
Workflow Guides¶
Use the CLI reference for argument meanings and defaults. Use the workflow guides for complete serving runs: