Avocet by Circuit Forge LLC — email classifier training tool: multi-account IMAP fetch, card-stack labeling UI, benchmark harness
- deberta-small: batch_size 16→8 + grad_accum 1→2 (same effective batch), gradient_checkpointing=True (fp16 stays off: DeBERTa v3 disentangled attention overflows fp16 at the gather step) - api: _best_cuda_device() picks highest free-VRAM GPU via nvidia-smi; sets CUDA_VISIBLE_DEVICES in subprocess env to prevent DataParallel replication across both GPUs; adds PYTORCH_ALLOC_CONF=expandable_segments - SSE log now reports which GPU was selected |
||
|---|---|---|
| app | ||
| config | ||
| data | ||
| docs | ||
| scripts | ||
| tests | ||
| web | ||
| .gitignore | ||
| CLAUDE.md | ||
| environment.yml | ||
| manage.sh | ||
| PRIVACY.md | ||
| pytest.ini | ||
| requirements.txt | ||