gitadmin/foxhunt
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

sp7(trainer): T5 polish — rename cql_dev → cql_sx_dev + base_wiener annotation

Browse Source 
Two clarity-only edits surfaced by code review:
- Rename `cql_dev` → `cql_sx_dev` in `launch_loss_balance_controller`.
  The launcher uses `grad_decomp_result_dev_ptr + 6 * f32_size` which
  is the cql_sx (post-budget) slot, not cql (pre-budget at offset 3).
  The variable name now reflects that.
- Add `// 213` inline annotation on `base_wiener_offset` to match the
  pattern used by all four sibling launchers (lines 11062, 11194, 11303).

No semantic impact — pointer arithmetic and parameter order unchanged.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This commit is contained in:
jgrusewski
2026-05-03 01:39:31 +02:00
parent ab6fbcb668
commit 40871595e8
2 changed files with 8 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
crates/ml/src/cuda_pipeline/gpu_dqn_trainer.rs
View File

@@ -11363,9 +11363,9 @@ impl GpuDqnTrainer {
// We use cql_sx (post-budget delta) for ratio parity with what landed
// in grad_buf.
let f32_size = std::mem::size_of::<f32>() as u64;
let iqn_dev = self.grad_decomp_result_dev_ptr + 0 * f32_size;
let cql_dev = self.grad_decomp_result_dev_ptr + 6 * f32_size;
let c51_dev = self.grad_decomp_result_dev_ptr + 9 * f32_size;
let iqn_dev = self.grad_decomp_result_dev_ptr + 0 * f32_size;
let cql_sx_dev = self.grad_decomp_result_dev_ptr + 6 * f32_size;
let c51_dev = self.grad_decomp_result_dev_ptr + 9 * f32_size;
// Step 1: producer kernel.
let flatness_isv_base_i32 = FLATNESS_BASE as i32;
@@ -11386,7 +11386,7 @@ impl GpuDqnTrainer {
self.stream
.launch_builder(&self.loss_balance_controller_kernel)
.arg(&iqn_dev)
.arg(&cql_dev)
.arg(&cql_sx_dev)
.arg(&c51_dev)
.arg(&isv_dev)
.arg(&flatness_isv_base_i32)
@@ -11413,7 +11413,7 @@ impl GpuDqnTrainer {
// Step 2: apply_pearls_ad_kernel × 24 — one per ISV output slot.
// Wiener offset: (SP4_PRODUCER_COUNT + (isv_slot - SP5_SLOT_BASE)) * 3.
let base_wiener_offset = SP4_PRODUCER_COUNT as i32 * 3;
let base_wiener_offset = SP4_PRODUCER_COUNT as i32 * 3; // 213
for (isv_base, scratch_base) in [
(BUDGET_CQL_BASE, SCRATCH_LB_BUDGET_CQL),

3
docs/dqn-wire-up-audit.md
View File

@@ -3901,6 +3901,9 @@ extended. No producer kernel yet — that arrives in the next commit.
constants + `launch_loss_balance_controller(&self)` fn (the producer
kernel launch + 24 apply_pearls_ad chain). Producer-only — call site
in training_loop.rs lands at T7 (atomic with consumer + stale doc).
T5 polish landed (commit ⟨pending⟩) — `cql_dev` renamed to `cql_sx_dev`
for clarity (the launcher uses the cql_sx post-budget delta) + `// 213`
annotation on `base_wiener_offset` for sibling-launcher consistency.
- T6: Pearl 2 contract change (drop CQL/C51/ENS args).
- T7: launch site + sentinel-aware bootstrap with bootstrap constants matching the kernel's cold-start basis (defined in T7) + stale doc.
- T9–T10: smoke + 50-epoch verification.