test: eps=2e-3 for flash dQ/dK finite-diff (cuts f32 rounding term)

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

crates/xtrain-autodiff/tests/autograd.rs

@@ -666,29 +666,25 @@ fn flash_attention_batched_bwd() {
         let (o, _) = qv.flash_attention(&kv, &vv, scale);
         weighted_sum(&o, &w)
     };
+    // Attention dQ/dK carry softmax curvature; for the small grad magnitudes here
+    // a larger eps (2e-3) cuts the f32 rounding term (∝|L|/eps) that dominates the
+    // O(eps²) truncation on a ~4e-4 grad. (dV is exactly linear → cfg_linear.)
+    let cfg_attn = GradCheckConfig {
+        eps: 2e-3,
+        rel_tol: 3e-2,
+        atol: 1e-3,
+    };
     let (kf, vf, ff) = (k_h.clone(), v_h.clone(), fwd.clone());
     let lq = move |x: &[f32], _s: &[usize]| ff(x, &kf, &vf);
     report(
         "flash dQ",
-        &grad_check(
-            &q_h,
-            &[bh, seq, hd],
-            &lq,
-            dq.as_slice::<f32>(),
-            cfg_nonlinear(),
-        ),
+        &grad_check(&q_h, &[bh, seq, hd], &lq, dq.as_slice::<f32>(), cfg_attn),
     );
     let (qf, vf, ff) = (q_h.clone(), v_h.clone(), fwd.clone());
     let lk = move |x: &[f32], _s: &[usize]| ff(&qf, x, &vf);
     report(
         "flash dK",
-        &grad_check(
-            &k_h,
-            &[bh, seq, hd],
-            &lk,
-            dk.as_slice::<f32>(),
-            cfg_nonlinear(),
-        ),
+        &grad_check(&k_h, &[bh, seq, hd], &lk, dk.as_slice::<f32>(), cfg_attn),
     );
     let (qf, kf, ff) = (q_h.clone(), k_h.clone(), fwd.clone());
     let lv = move |x: &[f32], _s: &[usize]| ff(&qf, &kf, x);