Rust tool that mmaps previous checkpoint, diffs against live GPU weights (via CUDA IPC handles), and only writes changed blocks. For small behavioral training steps, turns 54GB write into ~500MB. Also includes vllm_export_hook.py with direct source patch approach — exports IPC handles from vLLM's worker subprocess after model load. Run every 10 minutes via cron to protect against vLLM crashes. Daily rsync to moria for long-term storage. |
||
|---|---|---|
| .. | ||
| checkpoint | ||
| apollo_mini.py | ||
| apollo_worker.py | ||
| DESIGN.md | ||
| export_weights.py | ||
| start_vllm_with_apollo.sh | ||
| train.py | ||
| training_example.py | ||
| vllm_export_hook.py | ||
| weight_mapping.py | ||