ViperEkura a6c6a54ace feat: read host training vars from YAML infra section ...

- scripts/train.sh load_infra() parses the top-level infra: section of TRAIN_CONFIG_FILE
- exports TRAIN_JOB_NAME/DATA/MODEL/CHECKPOINT_DIR/TRAIN_GPU_COUNT/CUDA_VISIBLE_DEVICES
- infra overrides .env.train via compose interpolation precedence; keys absent fall back
- train.yaml is now the single per-job config: host mounts, GPU filter, and hyperparameters
- requires host python3 with PyYAML when TRAIN_CONFIG_FILE is set; errors fail fast
- docs: docker-training.md documents the infra overrides and precedence

2026-08-19 01:27:57 +08:00

..

architecture.md

docs: explain extension layer boundaries

2026-08-16 21:30:03 +08:00

cuda_kernels.md

perf: precompute ragged Q tile scheduling

2026-08-16 23:32:46 +08:00

dataflow.md

docs: sync all documentation with current codebase

2026-08-07 23:52:27 +08:00

docker-training.md

feat: read host training vars from YAML infra section

2026-08-19 01:27:57 +08:00

internals.md

docs: explain extension layer boundaries

2026-08-16 21:30:03 +08:00