Seth 3580d350b4 Parallel 3-GPU self-play: all tiers run simultaneously ...

Each cycle runs all three tiers at the same time on different GPUs:
- Tier 1 (drills) on GPU A
- Tier 2 (self-critique) on GPU B
- Tier 3 (adversarial) on GPU C
GPU assignments rotate each cycle for even wear.
3x throughput vs sequential. RCON handles concurrent commands.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

2026-03-20 00:55:24 -04:00

..

.gitkeep

Initial project scaffold: dataset schema, 31 seed training examples, Mineflayer bot framework, and 7-phase roadmap

2026-03-18 01:51:28 -04:00

distill.py

Swarm bots, RCON validation, Haiku distillation complete

2026-03-18 19:18:19 -04:00

generate_tool_training.py

Tool-calling training: 1,159 multi-turn examples with error correction

2026-03-19 18:49:08 -04:00

overnight_selfplay.sh

Parallel 3-GPU self-play: all tiers run simultaneously

2026-03-20 00:55:24 -04:00

self_play.py

3-tier self-play: command drills, self-critique, adversarial

2026-03-19 19:39:33 -04:00

train_lora.py

Mortdecai v4 pre-training: /no_think, dedup, 3,369 examples

2026-03-19 20:15:00 -04:00

train_lora.sh

Add LoRA training scripts and fix bake-off token budget

2026-03-18 10:40:18 -04:00

validate_distilled.py

Swarm bots, RCON validation, Haiku distillation complete

2026-03-18 19:18:19 -04:00