coolguy/coorl-lost-cities
1
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity
81 Commits 3 Branches 0 Tags
d0316293e410a002812102600bbcfd7ca9ec8847
Commit Graph
15 Commits
Author SHA1 Message Date
coolguyandClaude Haiku 4.5 d0316293e4 Document opponent_policy=network divergence finding 
opponent_policy=network 설정으로 1000-iter 실험 두 개 (512x3, 1024x4)를
돌린 결과 두 실험 모두 policy collapse가 발생함을 확인. 큰 capacity는
plateau를 늘리지만 발산 자체를 막지 못함. 향후 학습은 self_play_league
기본값을 유지할 것을 권고.

- 실험에 사용한 config 두 개 추가
- 발견 분석 문서 추가 (원인, 비교, 권고)

Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
 2026-05-07 13:47:36 +09:00
coolguyandClaude Haiku 4.5 cf67bc2198 Add color_shared + attention 1000-iteration experimental config 
- hidden_size: 256 (smaller for faster iteration)
- num_layers: 2
- color_attention_layers: 2, color_attention_heads: 8
- max_iterations: 1000
- advantage/strategy_updates: 256 (reduced from 512)
- Ready for experimental training run

Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
 2026-05-07 09:21:50 +09:00
coolguyandClaude Haiku 4.5 b7fce79f5b Add example configs for color_shared network architectures 
- Add deep_cfr_color_shared_512x3.yaml: color_shared without attention
- Add deep_cfr_color_shared_attention_512x3.yaml: color_shared with 2-layer attention
- Both use 512 hidden size and 3 MLP layers
- Configs are ready for experimental training runs

Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
 2026-05-07 09:19:02 +09:00
coolguyandClaude Haiku 4.5 fd99d3bb4a Deep CFR self-play anchor safe 512x3 2x updates 10000 iter config 및 관련 변경 
Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
 2026-05-07 09:14:23 +09:00
coolguy c779621823 Deep CFR pure self-play 512x3 2x updates config 추가 2026-05-07 07:55:05 +09:00
coolguy 55f4ae89a4 Deep CFR 512x3 LCFR config 추가 2026-05-07 07:12:39 +09:00
coolguy ccf7f825f7 Deep CFR 512x3 2x updates config 추가 2026-05-07 07:11:02 +09:00
coolguy 768c071e90 Deep CFR 512x3 unbounded config 추가 2026-05-07 07:05:11 +09:00
coolguy 5fa4bd65ef Deep CFR 무제한 학습 config 추가 2026-05-07 03:52:59 +09:00
coolguy 148be6e9a0 Deep CFR evaluation 배칭 및 병렬화 추가 2026-05-07 03:42:00 +09:00
coolguy 1b6d98ceeb Deep CFR traversal 운영 로그 보강 2026-05-07 01:23:26 +09:00
coolguy d29b3d60bb Deep CFR 재현 config 이름 정리 2026-05-07 01:15:34 +09:00
coolguy bbf8950c3a Deep CFR 재현 config run dir 정리 2026-05-07 01:05:23 +09:00
coolguy 9bcc88c1be Deep CFR legacy 재현 config 기반 추가 2026-05-07 00:46:09 +09:00
coolguy f593a6d910 Deep CFR YAML config 추가 2026-05-07 00:30:42 +09:00