A) TD-propagation diagnostic smoke test (sparse rewards)
- New smoke_tests/td_propagation.rs captures training_sharpe_ema across
20 epochs with micro_reward_scale=0 to answer: "can sparse-reward
Q-learning extract policy from trade-completion P&L alone?"
- Asserts: finite sharpe_ema, no monotonic degradation (tolerance 0.1
over first-3 vs last-3 epoch avg), q_gap > 0.05
- First run answered YES: val Sharpe peaks at +12.49 at epoch 8 with
pure sparse rewards, confirming the objective is learnable. The
remaining issue is stability/overfitting, not TD propagation.
B) Entropy → plasticity gate in training_loop
- D3/N3 shrink_perturb trigger now fires on `last_action_entropy < 0.3`
OR `health_value < 0.3` (OR semantics, 3 consecutive epochs).
- Catches action-collapse cases the generic health metric misses: Q-values
separate cleanly but argmax stays pinned to a single branch.
- tracing::info now logs both signals.
C) commitment_lambda coverage verified
- Almgren-Chriss sqrt market-impact already in compute_tx_cost
(trade_physics.cuh:168-171). commitment_lambda was pure duplication.
- Inventory doc updated: P2 marked satisfied, table row status updated.
Files touched:
- crates/ml/src/trainers/dqn/smoke_tests/mod.rs (+2)
- crates/ml/src/trainers/dqn/smoke_tests/td_propagation.rs (new)
- crates/ml/src/trainers/dqn/trainer/training_loop.rs (+15/-3)
- docs/superpowers/specs/2026-04-21-phase1-reward-inventory.md (+6/-3)
Smoke test PASSED locally (RTX 3050 Ti, 30.99s end-to-end).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Phase 2 P1 relocations — the intents behind two deleted reward shaping
terms now live at the correct layer: diagnostic monitoring, not reward.
1. training_sharpe_ema (was w_dsr reward input):
Field already existed. Added to HEALTH_DIAG log line as `sharpe_ema`.
Available for early-stopping, model selection, and exploration gating
without perturbing the objective.
2. action_entropy (was position_entropy_weight reward):
Computed from summary.action_counts at GPU summary download —
Shannon entropy normalized to [0, 1] by ln(n_bins). One-epoch lag
is fine for a diagnostic. Added to HEALTH_DIAG as `action_entropy`.
Stored in new trainer field `last_action_entropy: Option<f32>`.
New HEALTH_DIAG suffix: `diag [sharpe_ema={:.3} action_entropy={:.2}]`.
Verified visible on E1 smoke test epoch 18/19 output.
Also: honest recalibration of Phase 2 results added to the inventory
doc. The earlier commit messages framed "-150 → -17 Sharpe" as a 5×
improvement; in absolute terms Sharpe -17 is still catastrophic (you'd
blow the account). Phase 2 stopped active capital destruction but did
not find alpha. Sharpe_raw per bar went from -0.39 (lot of loss per bar)
to -0.09 (little loss per bar) — still net losing. q_gap=0.17 is a
capacity metric (network CAN differentiate), not profitability.
Path to actual profitability (Sharpe > 0) remains:
- TD propagation verification (Task #8)
- Unified env kernel (Phase 3)
- Possibly a different objective entirely
- Richer features (42 market + 20 OFI may be information-starved)
The rule going forward: before celebrating future improvements, ask
whether the result *crosses zero* (profitable) or is just *less
negative* (still losing).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Mass deletion of the "v7 gem" reward terms identified in the Phase 1
inventory as category errors — each one rewarded an outcome-adjacent
behavior instead of encoding the underlying physics, and each
empirically hurt validation metrics more than it helped training
stability.
Deleted from experience_kernels.cu and all plumbing (Rust configs,
launch args, hyperopt logs, TOML entries):
* order_credit_weight - reward redundant with compute_tx_cost
(order_type_idx already differentiates
fills by order type)
* risk_efficiency_weight - reward double-counted drawdown penalty
asymmetrically (only on winners)
* urgency_credit_weight - reward was vol-normalized unrealized P&L,
pure rename of core return
* commitment_lambda - triple-counted churn + tx_cost
* w_dsr - kernel wrote DSR EMA but no longer added
to reward (dead); removed the EMA
bookkeeping too
* dsr_eta - kernel arg for the deleted DSR EMA
* position_entropy_weight - rewarded action-bucket diversity
regardless of outcome; histogram buffer
+ zero-init removed too
* exit_timing_weight - already inactive (used raw_next future
price, comment-deleted earlier)
* ofi_reward_weight - dead plumbing; OFI already passed as
feature through state[OFI_START..]
* opportunity_cost_scale - penalized flat when Q-gap wide;
redundant with Q-values themselves
Kernel arg count: experience_env_step_batch shrank from ~55 to ~45 args.
Rust-side config surface reduced correspondingly.
Results on E1 smoke test (20-epoch):
BEFORE any Phase 2 work:
Val Sharpe -120 to -150, MaxDD 10-15%, Sharpe_raw -0.39
AFTER reward_noise + Kelly (both envs) + urgency + this sweep:
Val Sharpe -17 to -22 (7× better)
Val MaxDD 0.27% (40× better)
Val Sharpe_raw ~-0.09 (4× better)
Training Sharpe_raw ~0 (stabilized from ±20 swings)
Final q_gap 0.1712 (highest yet, collapse mechanism fine)
The extreme train-Sharpe swings (+17 one epoch, -13 next) were not
learning dynamics — they were shaping-term noise. Core reward (P&L +
drawdown + churn + holding + tx_cost + Kelly physics cap) gives training
metrics that actually reflect what the model does.
Inventory doc (docs/superpowers/specs/2026-04-21-phase1-reward-inventory.md)
extended with a "better-form taxonomy" section: every deleted gem has
a correct layer it belongs to (physics, feature, diagnostic, gradient-
level regularization — not reward). Kelly cap and Q-target smoothing
are already relocated; others are scheduled per the taxonomy's P1/P2/P3
priority list.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Extends the reward inventory with a review pass that preserves the
*ideas* behind behavioral terms while deleting the wrong-level mechanics.
Gems (relocate, don't delete):
- Probabilistic fill model (real physics, not shaping) → apply in both
modes via shared Philox RNG + exploration_scale
- Kelly-as-physics-constraint → hard position-size cap in trade_physics,
not a soft reward
- Path quality via per-step drawdown integration → no separate term,
just run drawdown penalty every bar
- Saboteur relocation → both modes, scaled by exploration_scale (training
1.0, validation 0.5, diagnostic 0.0) — avoids creating a new clean-vs-
noisy mismatch in the opposite direction
Pearls:
- Sparse reward is fine if TD propagation works — diagnose before
deleting micro_reward_scale; measure cov(Q(s_entry,a), trade_return)
- Every behavioral term maps to one of four failure modes: double-count,
misplaced physics, wrong-level regularization, or compensating for a
downstream bug. None are semantically "neutral" shaping.
Novel architectural moves:
- Diagnostic-only entropy tracking (no reward, just HEALTH_DIAG
logging)
- Single exploration_scale scalar replacing all mode toggles
(feedback_no_feature_flags compliant)
- Layered reward with per-term budget caps — makes the train/val Sharpe
gap computable instead of uncomputable magic
- Q-target smoothing replacing reward_noise_scale (regularization at
gradient level, not reward level)
Updated disposition table: 4 DELETE, 2 MOVE, 2 RELOCATE-to-physics,
1 DIAGNOSE-then-DELETE, 4 delete-dead-plumbing.
Three relocations (Kelly cap, saboteur scaling, Q-target smoothing) can
land as independent PRs before the unified_env_kernel rewrite, shrinking
Phase 2's change surface.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Enumerates every term that contributes to training or validation reward,
classifies each as P&L-aligned (keeper) or behavioral (remover), with
source line references into experience_kernels.cu and backtest_env_kernel.cu.
Findings:
- 7 P&L-aligned terms to preserve in the unified env (core return,
tx_cost, drawdown, inventory, churn, capital floor, + vol normalization)
- 8 active behavioral training-only terms to delete:
* order_credit_weight (rewards theoretical limit-order savings,
double-counts actual tx_cost)
* risk_efficiency_weight (double-counts drawdown asymmetrically)
* urgency_credit_weight (redundant with core return)
* kelly_sizing_weight (rewards matching a formula, not outcomes)
* micro_reward_scale (OFI-momentum signal follower)
* commitment_lambda (triple-counts churn)
* reward_noise_scale (belongs at gradient level, not reward level)
* position_entropy_weight (changes what "optimal" means)
- 4 dormant/dead terms to clean up (w_dsr, exit_timing, ofi_reward,
opp_cost_scale — all plumbing with no kernel effect)
Biggest offenders by magnitude: urgency_credit and risk_efficiency can
dominate the core signal on a single positioned bar.
Phase 2 (unified_env_kernel implementation) can now proceed with a
concrete deletion list, not a judgment call.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Brings the 4-commit work from wip/adaptive-learning-rootcause into main:
24f96ab78 fix(n1): distillation now actually pulls weights during collapse
f21a7d661 test(e1): 20-epoch smoke test + raw-q_gap assertion + disable early-stop
a3fd02a95 fix(early-stop): log real training epoch, not internal call counter
9dbd8d7e9 docs(design): unify training and validation environments
Key outcomes:
- Distillation mechanism now actually pulls weights during collapse (fixed
three bugs: epoch-boundary grad_buf erasure, CUDA-graph scalar baking,
wrong q_gap signal at snapshot gate). E1 smoke test passes deterministically.
- Design doc lays out the next step: unifying training and validation env
kernels so validation Sharpe tracks training Sharpe (currently diverges
catastrophically by ~-150 absolute).
- Incidental: early-stopping log now reports the real training epoch
instead of its internal call counter.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Design document for the follow-up to the adaptive-learning-rootcause
session. Lays out the evidence, root cause, options considered, and
recommended path for closing the 70% structural gap between training
and validation Sharpe that remained after the distillation collapse
fix landed.
Key findings documented:
- Reward-shaping ablation (2026-04-20) closed ~30% of the gap; ~70%
remains architectural
- Two env kernels (experience_env_step vs backtest_env_step) have
drifted: spread scaling, fill model, saboteur noise, reward terms,
action selection, position dynamics all differ
- Hint from history: experience_kernels.cu:1418 comment "Regime-
adaptive scaling removed to eliminate train/eval mismatch" shows
someone aligned *some* things previously
Recommended path (Option C, "unified env with layered reward"):
- Single unified_env_step kernel replaces both
- Core reward = pure P&L; shaping is additive and P&L-units-aligned
- Validation = training with exploration_scale=0 AND shaping_scale=0
- Scale factors are pinned device-mapped scalars (same pattern used by
the distillation alpha fix)
Phased implementation plan with ~8-day budget and concrete success
criteria: validation Sharpe_raw within 0.05 of training Sharpe_raw by
epoch 30 on L40S production run.
Rejected alternatives: backtest-matches-training (hides real issue),
training-matches-backtest (regresses stability), two-environment with
divergence as metric (fallback only).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Per user directive, ran rg scans for PINMEM, ROMEM, LOCKHOT, BORROW,
CPURO and populated the scoreboard with concrete file:line findings.
PINMEM: 10 new findings (PINMEM-011..019, plus rescan confirmed 001-010)
- Biggest surface: gpu_experience_collector.rs (7 htod sites)
- High priority: gpu_iqn_head.rs dtod_copy for iqn_rewards/iqn_dones
(PINMEM-013, score 25.0, switch to async-dtod, E=1)
ROMEM: 5 new findings (004 expanded, 005-008 added)
- ROMEM-004 now covers ~40 cuBLAS/cuBLASLt/cuDNN workspace casts across
shared_cublas_handle.rs, gpu_iql_trainer.rs, gpu_iqn_head.rs,
gpu_curiosity_trainer.rs, cublaslt_debug.rs — bulk false-positive
candidates (FFI convention, not actual RO writes)
- ROMEM-007 adds 6 more device-mapped pinned write sites (same pattern
as ROMEM-001/002 — benign cuMemHostAllocMapped)
LOCKHOT: 5 new findings (006-010)
- LOCKHOT-006/008: tokio::sync::Mutex<PPO> and RwLock<TLOBTransformer>
held across .await — deadlock risk. High priority.
- LOCKHOT-007: Arc<Mutex<VecDeque<f64>>> history locks per step
BORROW: 1 new finding (002)
- BORROW-002: RefCell<PPO> × 2 in validation/ppo_adapter.rs —
documented single-threaded, needs invariant verified
CPURO: deferred — .len()/.shape() scan returns hundreds of mostly-Vec
matches; left CPURO-000 task for next iter to classify per-site.
Scoreboard now has ~40 open findings across 12 active categories.
Expanding the scanner surface per user directive "all should be
addressed". Four distinct root causes for CPU-side violations around
GPU-owned data — each with its own scan command, fix hierarchy, and
learned-patterns entry.
- CPURO: CPU reads of GPU-resident data (sizes, reductions) that
force implicit sync. Cache at construction, keep stats on device.
- ROMEM: `*(ptr as *mut T)` writes where ptr came from a const /
read-only source. CUDA mapped-memory flags matter; cuBLAS/cuDNN
workspace casts are benign FFI.
- LOCKHOT: Mutex/RwLock on per-step path. High-value hits already
visible: Mutex<GpuDropout>, Mutex<Option<DropoutScheduler>> in
network.rs (every forward pass), Arc<Mutex<NStepBuffer>> in dqn.rs.
Never hold tokio::sync::RwLock across .await.
- BORROW: shared-&T promoted to &mut T via unsafe ptr casts or
UnsafeCell/RefCell. Real example shipped: gpu_replay_buffer.rs:690
changes CudaSlice<i32> → CudaSlice<u32> through raw-ptr cast.
Scoreboard seeded with 10 findings. Top scores:
- ROMEM-001 (25.0) - size_pinned mapped write, verify allocation flag
- ROMEM-004 (25.0) - cuBLAS workspace casts, likely false-positive bulk
- ROMEM-002 (15.0) - init-time pinned mapped writes
- LOCKHOT-001/002/003 (8.3) - dropout + nstep_buffer locks on hot path
- BORROW-001 (8.3) - CudaSlice element-type aliasing
CPURO is seeded with a scan-task (CPURO-000) to populate per-site
findings in iter 9 — too many Vec::len() false positives to list upfront.
Per user directive: htod/dtod transfers on the training hot path that
bounce through pageable host memory are a top-priority cleanup target.
Pinned memory (cuMemHostAlloc) + memcpy_htod_async enables DMA + stream
overlap. The codebase already has the pattern (size_pinned,
rng_step_pinned in gpu_replay_buffer.rs) — extend it to every per-step
transfer.
Prompt changes:
- New PINMEM category in §3 table (severity 5)
- Scan command excludes tests/benches/examples/smoke_tests
- New §6 learned pattern entry with 3-tier fix hierarchy:
(1) eliminate via on-device compute → (2) stage through pinned →
(3) async for dtod
- Added PINMEM to category ID allowlist in §2
Scoreboard seed: 10 initial findings from an iter-9 scan. Top-scoring:
PINMEM-003 (scratch_f32 scalar, E=1, score=25) and PINMEM-004 (init
scalars, score=15). Hot-path candidates include NoisyNet epsilon
refresh, PER replay-buffer inserts, target-net sync, PPO rollout.
Per user directive from iter 7 ("no deferred tasks, solve properly"),
the Needs-human review escape hatch is closed. Updated:
- Step 6c: replaced "move to Needs human review" with split-into-
sub-findings guidance; every finding must be solved.
- Step 6e: cargo-check failure now requires diagnosis + retry, not
deferral. Pre-existing external errors go to Known external state.
- Step 8 summary line: dropped the "deferred j" counter.
- Rule 7: similar rephrasing.
- New §6: learned patterns from iters 1-8 (dead flag chains,
aspirational config, dishonest fallbacks, inference-side flag flips,
accounting-only values, orphan files, hidden GPU sync).
Ralph re-feeds this file verbatim, so the next iteration picks up the
updated rules automatically.
- Layer 2 labeled as G2-G5, G5 explicit for gradient budget
- P2 spectral_gap_norm added as 7th component in composition (rebalanced weights)
- N6 ensemble oracle has explicit threshold (>0.8 triggers N3 immediately)
- N8 meta-Q wiring clarified: logged but not in composition until validated
- Files Changed deduplicated, paths qualified
- Success criteria: WinRate >55% (above random baseline ~25%)
- HEALTH_DIAG log line includes all 7 components
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Fixes Q-value collapse via unified LearningHealth signal that senses
training health (6 components) and continuously adapts:
- CQL regularization (regime + health gated)
- Gradient budget (IQN/CQL/Ens/C51 dynamic allocation)
- Tau target EMA (health-coupled)
- Expected SARSA temperature (continuous, no hardcoded threshold)
Plus 4 pearls (PER priorities, spectral detection, gradient consistency,
adaptive gamma) and 8 novels (self-distillation, barrier loss, plasticity
injection, CF curriculum, information bottleneck, ensemble oracle,
contrarian override, meta-Q network).
Core principle: training hyperparameters are OUTPUTS of the temporal
pipeline, not static schedules. The system meta-learns its own settings.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Training and validation use different kernels with different state layouts,
making generalization impossible. Single state_layout.cuh + one assembly
function + STATE_DIM as compile-time constant.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Removed the legacy non-branching Sequential q_network and target_network
from DQNAgent. These were never used (branching+dueling always active)
but allocated VRAM and ran noise resets every step. -190 lines.
Config tuning for dense micro-reward system:
- n_steps: 5→1 (TD(0), micro-rewards cancel over n>1)
- tau: 0.007→0.01 (faster target tracking for TD(0))
- c51_alpha_max: 0.5→1.0 (full C51, PopArt handles normalization)
- curiosity_weight: 0.1→0.0 (dense micro-reward replaces curiosity)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
target_dim expansion: adds raw_open (OHLCV) and mid_price_open
(MBP-10 midpoint at bar formation) to fxcache targets. FXCACHE_VERSION
2→3 for auto-rebuild. Legacy v2 files handled with close-price fallback.
Spec v5 adds 3 pearls:
- Bar duration encoding in Mamba2 (continuous-time SSM awareness)
- Order book center of mass from all 10 MBP-10 levels (aggression signal)
- Retrospective hold quality bonus (teaches exit timing)
Plus: Hold action (4th direction), DSR Sharpe EMA fix, counterfactual
magnitude/order sign fix, MFT mid-price mark-to-market.
OFI embed MLP now 18→10 (was 16→8). Mamba2 width SH2+10 (was SH2+8).
Attention width D+10 (was D+8).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
target_dim 4→6: adds raw_open (OHLCV) and mid_price_open (MBP-10
best bid/ask midpoint at bar formation) to the targets buffer.
Micro-reward now uses real intra-bar move (close - open) / atr
instead of close-to-close approximation. Also adds spread cost
awareness via |open - mid| penalty. Requires fxcache recomputation.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Major changes from v1:
- Fix OFI data pipeline (ofi_gpu never wired, indices 42→66)
- Expand PORTFOLIO_STRIDE 30→38 for prev-OFI storage
- Add Mamba2 d_h_history backward (2 new cuBLAS GEMMs)
- Expose attention d_input_scratch for gradient flow
- Pre-compute OFI deltas in experience collection (state[74..82])
- Lower rank normalization threshold to 1e-5 for micro-rewards
- Use close-to-close return instead of unavailable open_price
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Dense per-bar reward using order flow momentum × price confirmation.
OFI embedding MLP (16→8 via cuBLAS) feeds into Mamba2 history AND
attention input. Replaces sparse exit-only reward with continuous
temporal signal for the SSM and attention heads to learn from.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Single cuGraphExecLaunch per step. PER sampling, gather, training,
priority update ALL as child graph nodes in one parent. Direct-to-trainer
gather eliminates DtoD copies. GPU-side counters eliminate host writes.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Root cause: CUfunction sharing between graphed children and ungraphed
outside-graph ops corrupts kernel state on Hopper → 3100ms adam replay.
Fix: capture EVERYTHING unconditionally in child graphs. Selectivity,
denoise, causal intervention, vaccine, Q-stats all become graphed children.
Zero ungraphed launches = zero CUfunction conflicts.
7 children, ~190 kernel nodes, one capture, one replay per step.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Phase 1 results: 4.0s/step GPU compute (graph replay = ungraphed, cuBLAS
replaced manual matmul but feature count increased). CPU pipeline fully
async (0.1ms/step). Total ~718s/epoch — 100% GPU bound.
Updated Phase 2 targets: aux parallelism (2b) is highest priority —
80/177 kernels with 4 independent trainers. nsys profiling should run
first to validate SM vs memory-bound assumptions.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Inventory penalty makes flat optimal — model learns to do nothing.
Instead: penalize inaction proportional to model's own predicted edge
(Q-value spread). Creates virtuous cycle: better temporal attention →
higher self-imposed penalty for missed trades → more trading on signal.
No hindsight bias (uses predicted edge, not actual price change).
Micro-reward (already exists) rewards correct positioning.
Churn penalty (new) prevents rapid flips.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
forward_child: 4 branch streams fork after h_s2, join before loss
aux_child: IQL/IQN/attention on parallel streams
Uses CU_STREAM_CAPTURE_MODE_RELAXED with fork-join events as graph edges
Existing branch_streams[4] + events in batched_forward.rs ready to activate
Target: <40s epochs (from <80s Phase 1)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
