config(dqn): strip H100-tuned VRAM overrides from dqn-production.toml

dqn-production.toml hard-coded three H100-tuned values that shadow
GpuProfile auto-detection: batch_size=16384, buffer_size=500K,
gpu_n_episodes=4096. After the L40S-default flip lands (prior commit
8b8bb1af7), workflow `train-ft8ph` deterministically OOMed at fold 0/1/2
with `build_next_states_f32` 4 GiB alloc — because the H100-sized
hyperparams.batch_size + buffer_size + gpu_n_episodes ate ~38 GB of the
L40S's 46 GB usable VRAM before the rollout step.

DqnTrainingProfile.apply_to() runs AFTER train_baseline_rl.rs populates
hyperparams from GpuProfile, so the production TOML always wins. All
three fields are `Option<...>` in the TOML schema — removing the lines
turns apply_to into a no-op for them, and the GpuProfile-detected values
flow through:

  field            | L40S  | H100   | (was forced)
  batch_size       | 4096  | 8192   | 16384
  buffer_size      | 300K  | 500K   | 500K
  gpu_n_episodes   | 2048  | 4096   | 4096

Two pinned assertions in training_profile.rs::tests checked the old
contract `hp.batch_size == 16384`. Rewritten to assert
`hp.batch_size == baseline_batch_size` — locks the new contract that
VRAM-tuned values stay GpuProfile-sourced.

Lib suite 1016/0 green. Audit doc updated.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

config/training/dqn-production.toml

@@ -13,7 +13,11 @@
 
 [training]
 epochs = 200
-batch_size = 16384
+# batch_size: removed 2026-05-14 — sourced from `GpuProfile.training.batch_size`
+# (config/gpu/*.toml: H100=8192, L40S=4096, A100=2048, RTX3050=64). Hard-coded
+# 16384 here was H100-tuned and caused OOM on L40S after the 2026-05-14
+# argo gpu-pool default flip from H100 → L40S (workflow train-ft8ph: next_states
+# alloc 4 GiB collision at fold 0 from compounded H100-sized buffers).
 learning_rate = 1e-5
 gamma = 0.99
 weight_decay = 0.0001
@@ -52,7 +56,9 @@ q_gap_threshold = 0.05
 noise_sigma = 0.1
 
 [replay_buffer]
-buffer_size = 500000
+# buffer_size: removed 2026-05-14 — sourced from `GpuProfile.training.buffer_size`
+# (config/gpu/*.toml: H100=500K, L40S=300K, A100=200K, RTX3050=5K). Hard-coded
+# 500_000 here was H100-tuned. See [training].batch_size comment.
 min_replay_size = 1000
 per_alpha = 0.3  # was 0.6. Lower = more uniform. Dense micro-rewards have tiny TD-errors, high alpha ignores them.
 per_beta_start = 0.4
@@ -64,7 +70,12 @@ patience = 20
 min_epochs_before_stopping = 80
 
 [experience]
-gpu_n_episodes = 4096
+# gpu_n_episodes: removed 2026-05-14 — sourced from `GpuProfile.experience.
+# gpu_n_episodes` (config/gpu/*.toml: H100=4096, L40S=2048, A100=Some(?),
+# RTX3050=Some(?)). Hard-coded 4096 here was H100-tuned; on L40S the
+# `build_next_states_f32` rollout-step alloc = n_episodes × 2 × timesteps ×
+# state_dim × 4 = 4 GiB collides with the rest of the collector VRAM by the
+# rollout phase (workflow train-ft8ph repro). See [training].batch_size comment.
 initial_capital = 35000.0
 tx_cost_multiplier = 0.18  # bps: IBKR ES RT = $4.50/contract ($0.85 comm + $1.38 exch + $0.02 reg × 2 sides). At ES=$5100: 0.18 bps × 5100 × 0.0001 = 0.09 pts = $4.50
 

crates/ml/src/training_profile.rs

@@ -1198,12 +1198,18 @@ mod tests {
     #[test]
     fn test_apply_to_dqn_production() {
         let mut hp = crate::trainers::dqn::DQNHyperparameters::conservative();
+        let baseline_batch_size = hp.batch_size;
         let profile = DqnTrainingProfile::load("dqn-production");
         profile.apply_to(&mut hp);
 
         // [training] values must have been applied
         assert_eq!(hp.epochs, 200);
-        assert_eq!(hp.batch_size, 16384);
+        // batch_size: NOT overridden by dqn-production.toml (removed 2026-05-14
+        // — H100-tuned 16384 caused OOM on L40S default; the value now flows
+        // from `GpuProfile.training.batch_size`). Assertion: apply_to leaves
+        // `batch_size` at the conservative() baseline so the GpuProfile-supplied
+        // value (set earlier in train_baseline_rl.rs:443) flows through cleanly.
+        assert_eq!(hp.batch_size, baseline_batch_size);
         assert!((hp.learning_rate - 1e-5).abs() < 1e-10);
 
         // OFI MBP-10 data dir should be set in production profile
@@ -1468,11 +1474,15 @@ mod tests {
     fn test_production_profile_applies_all_sections() {
         let profile = DqnTrainingProfile::load("dqn-production");
         let mut hp = crate::trainers::dqn::DQNHyperparameters::conservative();
+        let baseline_batch_size = hp.batch_size;
         profile.apply_to(&mut hp);
 
         // [training] section
         assert_eq!(hp.epochs, 200);
-        assert_eq!(hp.batch_size, 16384);
+        // batch_size: see test_apply_to_dqn_production — VRAM-tuned values flow
+        // from GpuProfile, not the training TOML. apply_to leaves the conservative
+        // baseline in place so GpuProfile values aren't shadowed.
+        assert_eq!(hp.batch_size, baseline_batch_size);
         assert!((hp.learning_rate - 1e-5).abs() < 1e-10);
         assert!((hp.gamma - 0.99).abs() < 0.001);
         assert!((hp.reward_scale - 1.0).abs() < 0.01);

docs/dqn-wire-up-audit.md

@@ -2,6 +2,49 @@
 
 **Status:** Populated during Plan 1 Task 6 (A.5 orphan audit). Updated on every commit per Invariant 7.
 
+## 2026-05-14 — config(dqn): strip H100-tuned VRAM overrides from dqn-production.toml
+
+**Branch:** `sp20-aux-h-fixed`. **Trigger:** workflow `train-ft8ph` OOM at
+`build_next_states_f32` 4 GiB alloc on fold 0/1/2 after the L40S-default
+flip on the prior commit.
+
+**Root cause:** `config/training/dqn-production.toml` hard-coded three
+H100-tuned values that shadow `GpuProfile` auto-detection:
+
+| field | dqn-production | GpuProfile L40S | GpuProfile H100 |
+|-------|---------------:|----------------:|----------------:|
+| `batch_size`        | 16384  | 4096   | 8192   |
+| `buffer_size`       | 500K   | 300K   | 500K   |
+| `gpu_n_episodes`    | 4096   | 2048   | 4096   |
+
+The training profile loader's `apply_to(&mut hyperparams)` runs AFTER
+`train_baseline_rl.rs:443-490` populates hyperparams from `GpuProfile`
+— so the production TOML always wins. On L40S (46 GB usable, but only
+~7.86 GB free by the time the rollout-step `next_states` alloc fires
+because the H100-sized batch/buffer/episode buffers were already
+committed), the 4 GiB `n_episodes × 2 × timesteps × state_dim × 4 =
+4096 × 2 × 1000 × 128 × 4` allocation OOMs.
+
+**Fix:** delete the three lines from `dqn-production.toml`. All three
+fields are `Option<...>` in the TOML schema, so removal turns
+`apply_to` into a no-op for them; values then flow cleanly from
+`GpuProfile` (auto-detected per `profile_name_for_device("NVIDIA L40S")
+→ "l40s"`). H100 detection still gets H100 values; L40S gets L40S
+values; A100 / RTX3050 likewise.
+
+**Test updates:** two pinned assertions in `training_profile.rs::tests`
+(`test_apply_to_dqn_production` line 1206, `test_production_profile_
+applies_all_sections` line 1475) checked `hp.batch_size == 16384` after
+apply. Rewritten to check `hp.batch_size == baseline_batch_size` (i.e.,
+apply_to is a no-op on this field) — locks the new contract that VRAM-
+tuned values stay GpuProfile-sourced.
+
+**Tests:** all 19 `training_profile` tests pass. Full lib suite
+**1016/0 green** maintained.
+
+**Touched:** `config/training/dqn-production.toml`,
+`crates/ml/src/training_profile.rs` (test updates only).
+
 ## 2026-05-14 — SP22 H6 vNext: FoldReset registry entries for B5b/C collector buffers
 
 **Branch:** `sp20-aux-h-fixed`. **Builds on:** `4b40710b7` (B5b-2 collector