jgrusewski
31c610b69a
WIP: training quality fixes + VarStore elimination (INCOMPLETE — needs xavier_init_params_buf)
Phase 1 — Training quality (14 structural bugs):
- Reward: sqrt micro-reward scaling, remove sparse normalization, Kelly penalty,
additive OFI, magnitude-scaled capital floor penalty
- Gradient: CQL 25→10%, IQN 60→75%, MSE magnitude 4×, 3-phase counterfactual,
tau_anneal 100K→500K, epsilon floor 2%, PopArt fold reset, dd_threshold 0.5%
- Monitoring: per-branch Q-value instrumentation
- Hold: min_hold_bars 5→1
Phase 2 — VarStore elimination:
- Deleted copy_weights_from from all network types
- Deleted to_varstore from DuelingQNetwork + DistributionalDuelingQNetwork
- Deleted branching_to_varstore, 7 extract/sync VarStore functions
- Deleted update_target_networks, legacy CPU training paths
- Deleted iqn_target_network, target_network fields
- Deleted gpu_kernel_parity_test.rs (705 lines)
- Refactored GpuExperienceCollector to single flat buffer
- Added weight_sets_from_branching() zero-copy replacement
- EMA tau=1.0 init for target_params_buf
KNOWN ISSUE: params_buf is zero-initialized but from_flat_buffer() creates
weight set pointers into it BEFORE flatten_online_weights runs. The online
forward reads zeros → NaN. Needs xavier_init_params_buf() at construction.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-04-11 21:03:05 +02:00
..
2026-03-15 12:00:13 +01:00
2026-03-15 12:00:13 +01:00
2026-03-15 12:00:13 +01:00
2026-03-15 12:00:13 +01:00
2026-03-15 12:00:13 +01:00
2026-03-15 12:00:13 +01:00
2026-03-15 12:00:13 +01:00
2026-04-11 11:54:09 +02:00
2026-03-29 21:32:04 +02:00
2026-03-15 12:00:13 +01:00
2026-03-13 10:30:51 +01:00
2026-03-15 12:00:13 +01:00
2026-03-13 10:18:35 +01:00
2026-04-11 11:54:09 +02:00
2026-04-11 11:54:09 +02:00
2026-03-15 12:00:13 +01:00
2026-04-11 11:54:09 +02:00
2026-03-13 10:18:35 +01:00
2026-03-19 10:02:26 +01:00
2026-04-11 11:54:09 +02:00
2026-04-11 11:54:09 +02:00
2026-03-13 10:18:35 +01:00
2026-03-13 10:18:35 +01:00
2026-04-11 11:54:09 +02:00
2026-04-11 11:54:09 +02:00
2026-03-19 10:02:26 +01:00
2026-04-04 10:02:37 +02:00
2026-04-04 01:04:45 +02:00
2026-03-13 10:18:35 +01:00
2026-04-11 11:54:09 +02:00
2026-04-11 11:54:09 +02:00
2026-03-15 12:00:13 +01:00
2026-03-15 12:00:13 +01:00
2026-03-15 12:00:13 +01:00
2026-03-15 12:00:13 +01:00
2026-03-19 10:02:26 +01:00
2026-03-19 08:55:55 +01:00
2026-03-15 12:00:13 +01:00
2026-03-19 10:02:26 +01:00
2026-04-04 10:02:37 +02:00
2026-03-15 12:00:13 +01:00
2026-03-15 12:00:13 +01:00
2026-04-01 01:13:19 +02:00
2026-03-25 09:45:54 +01:00
2026-04-04 10:02:37 +02:00
2026-04-11 11:54:09 +02:00
2026-04-11 11:54:09 +02:00
2026-03-13 10:18:35 +01:00
2026-03-31 01:30:33 +02:00
2026-03-31 01:30:33 +02:00
2026-03-13 10:18:35 +01:00
2026-04-04 10:02:37 +02:00
2026-04-11 11:54:09 +02:00
2026-03-19 10:02:26 +01:00
2026-04-04 10:02:37 +02:00
2026-04-11 11:54:09 +02:00
2026-03-16 21:01:28 +01:00
2026-03-16 21:01:28 +01:00
2026-03-15 12:00:13 +01:00
2026-03-15 12:00:13 +01:00
2026-03-13 10:18:35 +01:00
2026-04-11 15:21:21 +02:00
2026-03-31 10:56:20 +02:00
2026-04-04 15:25:02 +02:00
2026-04-11 11:54:09 +02:00
2026-04-11 11:54:09 +02:00
2026-03-25 09:45:54 +01:00
2026-03-29 21:32:04 +02:00
2026-03-13 10:18:35 +01:00
2026-03-15 12:00:13 +01:00
2026-03-31 01:30:33 +02:00
2026-04-11 11:54:09 +02:00
2026-03-13 10:18:35 +01:00
2026-03-15 12:00:13 +01:00
2026-03-15 12:00:13 +01:00
2026-04-10 18:45:49 +02:00
2026-04-10 18:52:21 +02:00
2026-04-11 15:21:21 +02:00
2026-03-25 09:45:54 +01:00
2026-03-19 10:02:26 +01:00
2026-03-31 10:56:20 +02:00
2026-03-13 10:18:35 +01:00
2026-03-31 01:30:33 +02:00
2026-03-13 10:18:35 +01:00
2026-03-13 10:18:35 +01:00
2026-03-15 12:00:13 +01:00
2026-03-13 10:18:35 +01:00
2026-03-13 10:18:35 +01:00
2026-03-13 10:18:35 +01:00
2026-03-19 10:02:26 +01:00
2026-03-13 10:18:35 +01:00
2026-03-19 10:02:26 +01:00
2026-03-13 10:18:35 +01:00
2026-03-15 12:00:13 +01:00
2026-03-15 12:00:13 +01:00
2026-03-13 10:18:35 +01:00
2026-03-19 08:59:52 +01:00
2026-03-20 15:41:42 +01:00
2026-03-19 10:02:26 +01:00
2026-03-18 00:53:47 +01:00
2026-03-13 10:18:35 +01:00
2026-03-19 10:02:26 +01:00
2026-04-01 01:13:19 +02:00
2026-03-15 12:00:13 +01:00
2026-03-13 10:18:35 +01:00
2026-03-13 10:18:35 +01:00
2026-03-15 12:00:13 +01:00
2026-03-31 01:30:33 +02:00
2026-03-15 12:00:13 +01:00
2026-03-15 12:00:13 +01:00
2026-03-15 12:00:13 +01:00
2026-04-11 11:54:09 +02:00
2026-04-11 11:54:09 +02:00
2026-04-11 11:54:09 +02:00
2026-04-11 11:54:09 +02:00
2026-04-11 11:54:09 +02:00
2026-04-11 11:54:09 +02:00
2026-04-11 11:54:09 +02:00
2026-03-19 10:02:26 +01:00
2026-03-19 08:59:52 +01:00
2026-03-13 10:18:35 +01:00
2026-03-13 10:18:35 +01:00
2026-03-13 10:18:35 +01:00
2026-03-13 10:18:35 +01:00
2026-03-19 10:02:26 +01:00
2026-04-11 15:21:21 +02:00
2026-03-19 10:02:26 +01:00
2026-03-19 10:02:26 +01:00
2026-03-15 12:00:13 +01:00
2026-03-15 12:00:13 +01:00
2026-03-19 10:02:26 +01:00
2026-04-11 11:54:09 +02:00
2026-03-13 10:18:35 +01:00
2026-03-13 10:18:35 +01:00
2026-04-11 11:54:09 +02:00
2026-04-02 14:21:45 +02:00
2026-03-13 10:18:35 +01:00
2026-04-11 11:54:09 +02:00
2026-03-19 08:59:52 +01:00
2026-03-15 12:00:13 +01:00
2026-03-13 10:18:35 +01:00
2026-03-15 12:00:13 +01:00
2026-03-19 10:02:26 +01:00
2026-03-19 10:02:26 +01:00
2026-03-19 10:02:26 +01:00
2026-04-11 11:54:09 +02:00
2026-03-13 10:18:35 +01:00
2026-04-04 10:02:37 +02:00
2026-03-13 10:18:35 +01:00
2026-03-13 10:18:35 +01:00
2026-03-19 10:02:26 +01:00
2026-03-15 12:00:13 +01:00
2026-03-13 10:18:35 +01:00
2026-03-19 08:55:55 +01:00
2026-03-13 10:18:35 +01:00
2026-03-15 12:00:13 +01:00
2026-04-11 11:54:09 +02:00
2026-03-13 10:18:35 +01:00
2026-03-13 10:18:35 +01:00
2026-03-13 10:18:35 +01:00
2026-03-19 10:02:26 +01:00
2026-04-11 21:03:05 +02:00
2026-03-19 10:02:26 +01:00
2026-03-19 10:02:26 +01:00
2026-03-21 09:14:53 +01:00
2026-04-10 18:52:21 +02:00
2026-03-19 08:59:52 +01:00
2026-03-19 10:02:26 +01:00
2026-03-15 12:00:13 +01:00
2026-03-19 08:44:50 +01:00
2026-03-19 10:02:26 +01:00
2026-03-15 12:00:13 +01:00
2026-03-15 12:00:13 +01:00
2026-03-19 10:02:26 +01:00
2026-03-19 10:02:26 +01:00
2026-03-13 10:18:35 +01:00
2026-03-19 08:44:50 +01:00
2026-03-15 12:00:13 +01:00
2026-03-19 10:02:26 +01:00
2026-03-19 10:02:26 +01:00
2026-03-19 08:44:50 +01:00
2026-03-19 08:59:52 +01:00
2026-03-16 16:11:48 +01:00
2026-03-14 11:35:15 +01:00
2026-04-11 15:21:21 +02:00
2026-04-11 11:54:09 +02:00
2026-03-13 10:18:35 +01:00
2026-03-15 12:00:13 +01:00
2026-03-15 12:00:13 +01:00
2026-03-15 12:00:13 +01:00
2026-04-04 10:02:37 +02:00
2026-04-11 11:54:09 +02:00
2026-04-11 11:54:09 +02:00
2026-03-15 12:00:13 +01:00
2026-03-15 12:00:13 +01:00
2026-03-13 10:18:35 +01:00
2026-03-19 10:02:26 +01:00