Update README.md

README.md

@@ -1,3 +1,169 @@
----
-license: apache-2.0
----
+---
+license: apache-2.0
+---
+
+First barrage tests and compilation tests are finished. The orbital is yielding a true representation of omega.
+
+
+```
+geolip_core.linalg: available
+geolip.linalg backend:
+  CUDA:       yes
+  Triton:     3.6.0
+  FL eigh:    enabled
+  Triton SVD: enabled
+  GPU:        NVIDIA RTX PRO 6000 Blackwell Server Edition
+========================================================================
+  Flow Ensemble — Expanded Test Suite
+========================================================================
+  device=cuda  geolip_core.linalg=True
+  GPU: NVIDIA RTX PRO 6000 Blackwell Server Edition
+
+========================================================================
+  1. SMOKE TEST
+========================================================================
+
+  Flow                     Params          Shape       Time     Conf   Res norm
+  ────────────────────── ──────── ────────────── ────────── ──────── ──────────
+  QuaternionFlow           70,725  (16, 64, 128)      264us    0.476      1.318
+  QuaternionLiteFlow       87,109  (16, 64, 128)      190us    0.465      1.279
+  VelocityFlow             70,210  (16, 64, 128)      170us    0.533      0.056
+  MagnitudeFlow            73,293  (16, 64, 128)    20.58ms    0.481      0.753
+  OrbitalFlow              25,445  (16, 64, 128)    20.70ms    0.551      3.464
+  AlignmentFlow            37,186  (16, 64, 128)    43.67ms    0.480      0.131
+
+========================================================================
+  2. LINALG INTEGRATION
+========================================================================
+
+  Testing eigh dispatch in MagnitudeFlow and OrbitalFlow...
+  magnitude          finite=True  conf=0.477
+  orbital            finite=True  conf=0.501
+
+  Gram eigenspectrum: shape=(16, 12) range=[0.0001, 6.6260]
+  Eigenvector orth err: 8.39e+01
+
+========================================================================
+  3. MULTI-SCALE
+========================================================================
+
+  OrbitalFlow across scales:
+  Config        B     n     k     d       Time   OK
+  ────────── ──── ───── ───── ───── ────────── ────
+  tiny          4    16     8    64    12.05ms   OK
+  small        16    64    32   128    20.64ms   OK
+  medium       32   128    64   256    20.71ms   OK
+  large        64   256   128   256    20.75ms   OK
+  wide          8   512   256   512    20.74ms   OK
+
+========================================================================
+  4. ENSEMBLE FUSION
+========================================================================
+
+  weighted:  time=87.28ms  norm=0.586  diversity=0.829
+    quaternion         conf=0.476±0.002  res=1.322
+    quat_lite          conf=0.465±0.002  res=1.285
+    velocity           conf=0.533±0.003  res=0.056
+    magnitude          conf=0.481±0.004  res=0.757
+    orbital            conf=0.554±0.002  res=1.644
+    alignment          conf=0.480±0.003  res=0.131
+
+  gated:  time=87.36ms  norm=0.598  diversity=0.829
+    quaternion         conf=0.476±0.002  res=1.322
+    quat_lite          conf=0.465±0.002  res=1.285
+    velocity           conf=0.533±0.003  res=0.056
+    magnitude          conf=0.481±0.004  res=0.757
+    orbital            conf=0.554±0.002  res=1.644
+    alignment          conf=0.480±0.003  res=0.131
+
+  residual:  time=87.35ms  norm=0.586  diversity=0.829
+    quaternion         conf=0.476±0.002  res=1.322
+    quat_lite          conf=0.465±0.002  res=1.285
+    velocity           conf=0.533±0.003  res=0.056
+    magnitude          conf=0.481±0.004  res=0.757
+    orbital            conf=0.554±0.002  res=1.644
+    alignment          conf=0.480±0.003  res=0.131
+
+========================================================================
+  5. GRADIENT HEALTH
+========================================================================
+
+  Flow               Loss          Grad norm   Status
+  ────────────────── ────────── ──────────── ────────
+  quaternion         mse            8.90e-04       OK
+  quat_lite          mse            7.05e-04       OK
+  velocity           mse            2.78e-04       OK
+  magnitude          mse            7.92e-04       OK
+  orbital            mse            3.45e-03       OK
+  alignment          mse            1.27e-03       OK
+  quaternion         cosine         1.43e-01       OK
+  quat_lite          cosine         1.17e-01       OK
+  velocity           cosine         3.37e-02       OK
+  magnitude          cosine         1.98e-01       OK
+  orbital            cosine         2.82e-01       OK
+  alignment          cosine         9.83e-02       OK
+  quaternion         norm           1.09e-01       OK
+  quat_lite          norm           8.79e-02       OK
+  velocity           norm           2.06e-02       OK
+  magnitude          norm           3.99e-01       OK
+  orbital            norm           3.97e+00       OK
+  alignment          norm           1.01e-01       OK
+
+========================================================================
+  6. ABLATION (100 training steps, rotation target)
+========================================================================
+
+  Configuration                              MSE     Params
+  ─────────────────────────────────── ────────── ──────────
+  QuaternionFlow                          0.0010    280,709
+  QuaternionLiteFlow                      0.0005    346,245
+  VelocityFlow                            0.0061    279,682
+  MagnitudeFlow                           0.0081    285,837
+  OrbitalFlow                             0.0456     91,813
+  AlignmentFlow                           0.0066    148,098
+  Quat + Orbital                          0.0024    372,524
+  Velocity + Magnitude                    0.0062    565,521
+  Orbital + Alignment                     0.0188    239,913
+  Velocity + Orbital                      0.1576    371,497
+  Full (weighted)                         0.0046  1,432,390
+/tmp/ipykernel_16934/1749400227.py:389: UserWarning: torch.linalg.svd: During SVD computation with the selected cusolver driver, batches 0, 1, 2, 3, 4, and other 27 batches failed to converge. A more accurate method will be used to compute the SVD as a fallback. Check doc at https://pytorch.org/docs/stable/generated/torch.linalg.svd.html (Triggered internally at /pytorch/aten/src/ATen/native/cuda/linalg/BatchLinearAlgebraLib.cpp:701.)
+  U, _, Vh = torch.linalg.svd(C)  # full d×d SVD — not through geolip.linalg.svd
+  Full (residual)                     FAILED: linalg.svd: (Batch element 0):
+
+========================================================================
+  7. COMPILE COMPATIBILITY
+========================================================================
+
+  Flow                      fullgraph        Raw     Compiled
+  ────────────────────── ──────────── ────────── ────────────
+  QuaternionFlow                   OK      269us        191us
+  QuaternionLiteFlow               OK      191us        172us
+  VelocityFlow                     OK      176us        161us
+/usr/local/lib/python3.12/dist-packages/torch/_inductor/compile_fx.py:321: UserWarning: TensorFloat32 tensor cores for float32 matrix multiplication available but not enabled. Consider setting `torch.set_float32_matmul_precision('high')` for better performance.
+  warnings.warn(
+/usr/local/lib/python3.12/dist-packages/torch/_inductor/lowering.py:1904: FutureWarning: `torch._prims_common.check` is deprecated and will be removed in the future. Please use `torch._check*` functions instead.
+  check(
+  MagnitudeFlow                    OK    20.68ms        733us
+/usr/local/lib/python3.12/dist-packages/torch/_inductor/lowering.py:1904: FutureWarning: `torch._prims_common.check` is deprecated and will be removed in the future. Please use `torch._check*` functions instead.
+  check(
+  OrbitalFlow                      OK    20.71ms        732us
+  AlignmentFlow                    OK     9.11ms       9.02ms
+
+========================================================================
+  8. MEMORY (B=32, n=128, k=64, d=256)
+========================================================================
+
+  Flow                      Peak MB
+  ────────────────────── ──────────
+  QuaternionFlow              36.2
+  QuaternionLiteFlow          28.3
+  VelocityFlow                48.0
+  MagnitudeFlow               24.2
+  OrbitalFlow                 10.5
+  AlignmentFlow               60.0
+  Full ensemble              203.6
+
+========================================================================
+  Done.
+========================================================================
+```