Update README.md
Browse files
README.md
CHANGED
|
@@ -5,21 +5,6 @@ license: apache-2.0
|
|
| 5 |
First barrage tests and compilation tests are finished. The orbital is yielding a true representation of omega.
|
| 6 |
|
| 7 |
The speed and accuracy is both representationally superior and compiles safe to full-graph.
|
| 8 |
-
|
| 9 |
-
```
|
| 10 |
-
Flow fullgraph Raw Compiled
|
| 11 |
-
ββββββββββββββββββββββ ββββββββββββ ββββββββββ ββββββββββββ
|
| 12 |
-
OrbitalFlow OK 20.71ms 732us
|
| 13 |
-
AlignmentFlow OK 9.11ms 9.02ms
|
| 14 |
-
|
| 15 |
-
Flow Peak MB
|
| 16 |
-
ββββββββββββββββββββββ ββββββββββ
|
| 17 |
-
OrbitalFlow 10.5
|
| 18 |
-
AlignmentFlow 60.0
|
| 19 |
-
Full ensemble 203.6
|
| 20 |
-
|
| 21 |
-
```
|
| 22 |
-
|
| 23 |
```
|
| 24 |
geolip_core.linalg: available
|
| 25 |
geolip.linalg backend:
|
|
@@ -40,23 +25,23 @@ geolip.linalg backend:
|
|
| 40 |
|
| 41 |
Flow Params Shape Time Conf Res norm
|
| 42 |
ββββββββββββββββββββββ ββββββββ ββββββββββββββ ββββββββββ ββββββββ ββββββββββ
|
| 43 |
-
QuaternionFlow 70,725 (16, 64, 128)
|
| 44 |
-
QuaternionLiteFlow 87,109 (16, 64, 128)
|
| 45 |
-
VelocityFlow 70,210 (16, 64, 128)
|
| 46 |
-
MagnitudeFlow 73,293 (16, 64, 128)
|
| 47 |
-
OrbitalFlow 25,445 (16, 64, 128)
|
| 48 |
-
AlignmentFlow
|
| 49 |
|
| 50 |
========================================================================
|
| 51 |
2. LINALG INTEGRATION
|
| 52 |
========================================================================
|
| 53 |
|
| 54 |
Testing eigh dispatch in MagnitudeFlow and OrbitalFlow...
|
| 55 |
-
magnitude finite=True conf=0.
|
| 56 |
-
orbital finite=True conf=0.
|
| 57 |
|
| 58 |
-
Gram eigenspectrum: shape=(16, 12) range=[0.
|
| 59 |
-
Eigenvector orth err:
|
| 60 |
|
| 61 |
========================================================================
|
| 62 |
3. MULTI-SCALE
|
|
@@ -65,39 +50,39 @@ geolip.linalg backend:
|
|
| 65 |
OrbitalFlow across scales:
|
| 66 |
Config B n k d Time OK
|
| 67 |
ββββββββββ ββββ βββββ βββββ βββββ ββββββββββ ββββ
|
| 68 |
-
tiny 4 16 8 64
|
| 69 |
-
small 16 64 32 128
|
| 70 |
-
medium 32 128 64 256
|
| 71 |
-
large 64 256 128 256
|
| 72 |
-
wide 8 512 256 512
|
| 73 |
|
| 74 |
========================================================================
|
| 75 |
4. ENSEMBLE FUSION
|
| 76 |
========================================================================
|
| 77 |
|
| 78 |
-
weighted: time=
|
| 79 |
-
quaternion conf=0.
|
| 80 |
-
quat_lite conf=0.
|
| 81 |
-
velocity conf=0.
|
| 82 |
-
magnitude conf=0.
|
| 83 |
-
orbital conf=0.
|
| 84 |
-
alignment conf=0.
|
| 85 |
|
| 86 |
-
gated: time=
|
| 87 |
-
quaternion conf=0.
|
| 88 |
-
quat_lite conf=0.
|
| 89 |
-
velocity conf=0.
|
| 90 |
-
magnitude conf=0.
|
| 91 |
-
orbital conf=0.
|
| 92 |
-
alignment conf=0.
|
| 93 |
|
| 94 |
-
residual: time=
|
| 95 |
-
quaternion conf=0.
|
| 96 |
-
quat_lite conf=0.
|
| 97 |
-
velocity conf=0.
|
| 98 |
-
magnitude conf=0.
|
| 99 |
-
orbital conf=0.
|
| 100 |
-
alignment conf=0.
|
| 101 |
|
| 102 |
========================================================================
|
| 103 |
5. GRADIENT HEALTH
|
|
@@ -106,23 +91,23 @@ geolip.linalg backend:
|
|
| 106 |
Flow Loss Grad norm Status
|
| 107 |
ββββββββββββββββββ ββββββββββ ββββββββββββ ββββββββ
|
| 108 |
quaternion mse 8.90e-04 OK
|
| 109 |
-
quat_lite mse
|
| 110 |
-
velocity mse
|
| 111 |
-
magnitude mse
|
| 112 |
-
orbital mse
|
| 113 |
-
alignment mse 1.
|
| 114 |
-
quaternion cosine
|
| 115 |
-
quat_lite cosine
|
| 116 |
-
velocity cosine
|
| 117 |
-
magnitude cosine
|
| 118 |
-
orbital cosine
|
| 119 |
-
alignment cosine
|
| 120 |
-
quaternion norm
|
| 121 |
-
quat_lite norm
|
| 122 |
-
velocity norm
|
| 123 |
-
magnitude norm
|
| 124 |
-
orbital norm
|
| 125 |
-
alignment norm
|
| 126 |
|
| 127 |
========================================================================
|
| 128 |
6. ABLATION (100 training steps, rotation target)
|
|
@@ -130,20 +115,18 @@ geolip.linalg backend:
|
|
| 130 |
|
| 131 |
Configuration MSE Params
|
| 132 |
βββββββββββββββββββββββββββββββββββ ββββββββββ ββββββββββ
|
| 133 |
-
QuaternionFlow 0.
|
| 134 |
QuaternionLiteFlow 0.0005 346,245
|
| 135 |
VelocityFlow 0.0061 279,682
|
| 136 |
MagnitudeFlow 0.0081 285,837
|
| 137 |
-
OrbitalFlow 0.
|
| 138 |
-
AlignmentFlow 0.
|
| 139 |
-
Quat + Orbital 0.
|
| 140 |
Velocity + Magnitude 0.0062 565,521
|
| 141 |
-
Orbital + Alignment 0.
|
| 142 |
-
Velocity + Orbital 0.
|
| 143 |
-
Full (weighted) 0.
|
| 144 |
-
|
| 145 |
-
U, _, Vh = torch.linalg.svd(C) # full dΓd SVD β not through geolip.linalg.svd
|
| 146 |
-
Full (residual) FAILED: linalg.svd: (Batch element 0):
|
| 147 |
|
| 148 |
========================================================================
|
| 149 |
7. COMPILE COMPATIBILITY
|
|
@@ -151,18 +134,12 @@ geolip.linalg backend:
|
|
| 151 |
|
| 152 |
Flow fullgraph Raw Compiled
|
| 153 |
ββββββββββββββββββββββ ββββββββββββ ββββββββββ ββββββββββββ
|
| 154 |
-
QuaternionFlow OK
|
| 155 |
-
QuaternionLiteFlow OK
|
| 156 |
-
VelocityFlow OK
|
| 157 |
-
|
| 158 |
-
|
| 159 |
-
|
| 160 |
-
check(
|
| 161 |
-
MagnitudeFlow OK 20.68ms 733us
|
| 162 |
-
/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.
|
| 163 |
-
check(
|
| 164 |
-
OrbitalFlow OK 20.71ms 732us
|
| 165 |
-
AlignmentFlow OK 9.11ms 9.02ms
|
| 166 |
|
| 167 |
========================================================================
|
| 168 |
8. MEMORY (B=32, n=128, k=64, d=256)
|
|
@@ -174,9 +151,9 @@ geolip.linalg backend:
|
|
| 174 |
QuaternionLiteFlow 28.3
|
| 175 |
VelocityFlow 48.0
|
| 176 |
MagnitudeFlow 24.2
|
| 177 |
-
OrbitalFlow 10.
|
| 178 |
-
AlignmentFlow
|
| 179 |
-
Full ensemble
|
| 180 |
|
| 181 |
========================================================================
|
| 182 |
Done.
|
|
|
|
| 5 |
First barrage tests and compilation tests are finished. The orbital is yielding a true representation of omega.
|
| 6 |
|
| 7 |
The speed and accuracy is both representationally superior and compiles safe to full-graph.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 8 |
```
|
| 9 |
geolip_core.linalg: available
|
| 10 |
geolip.linalg backend:
|
|
|
|
| 25 |
|
| 26 |
Flow Params Shape Time Conf Res norm
|
| 27 |
ββββββββββββββββββββββ ββββββββ ββββββββββββββ ββββββββββ ββββββββ ββββββββββ
|
| 28 |
+
QuaternionFlow 70,725 (16, 64, 128) 267us 0.489 1.355
|
| 29 |
+
QuaternionLiteFlow 87,109 (16, 64, 128) 194us 0.471 1.282
|
| 30 |
+
VelocityFlow 70,210 (16, 64, 128) 175us 0.495 0.058
|
| 31 |
+
MagnitudeFlow 73,293 (16, 64, 128) 315us 0.526 0.601
|
| 32 |
+
OrbitalFlow 25,445 (16, 64, 128) 378us 0.524 1.650
|
| 33 |
+
AlignmentFlow 8,922 (16, 64, 128) 347us 0.485 2.905
|
| 34 |
|
| 35 |
========================================================================
|
| 36 |
2. LINALG INTEGRATION
|
| 37 |
========================================================================
|
| 38 |
|
| 39 |
Testing eigh dispatch in MagnitudeFlow and OrbitalFlow...
|
| 40 |
+
magnitude finite=True conf=0.469
|
| 41 |
+
orbital finite=True conf=0.462
|
| 42 |
|
| 43 |
+
Gram eigenspectrum: shape=(16, 12) range=[0.0002, 7.5887]
|
| 44 |
+
Eigenvector orth err: 6.86e+01
|
| 45 |
|
| 46 |
========================================================================
|
| 47 |
3. MULTI-SCALE
|
|
|
|
| 50 |
OrbitalFlow across scales:
|
| 51 |
Config B n k d Time OK
|
| 52 |
ββββββββββ ββββ βββββ βββββ βββββ ββββββββββ ββββ
|
| 53 |
+
tiny 4 16 8 64 328us OK
|
| 54 |
+
small 16 64 32 128 377us OK
|
| 55 |
+
medium 32 128 64 256 379us OK
|
| 56 |
+
large 64 256 128 256 430us OK
|
| 57 |
+
wide 8 512 256 512 426us OK
|
| 58 |
|
| 59 |
========================================================================
|
| 60 |
4. ENSEMBLE FUSION
|
| 61 |
========================================================================
|
| 62 |
|
| 63 |
+
weighted: time=2.02ms norm=0.805 diversity=0.905
|
| 64 |
+
quaternion conf=0.489Β±0.002 res=1.355
|
| 65 |
+
quat_lite conf=0.471Β±0.002 res=1.285
|
| 66 |
+
velocity conf=0.495Β±0.003 res=0.058
|
| 67 |
+
magnitude conf=0.526Β±0.004 res=0.608
|
| 68 |
+
orbital conf=0.524Β±0.002 res=1.647
|
| 69 |
+
alignment conf=0.485Β±0.014 res=3.110
|
| 70 |
|
| 71 |
+
gated: time=2.02ms norm=0.663 diversity=0.905
|
| 72 |
+
quaternion conf=0.489Β±0.002 res=1.355
|
| 73 |
+
quat_lite conf=0.471Β±0.002 res=1.285
|
| 74 |
+
velocity conf=0.495Β±0.003 res=0.058
|
| 75 |
+
magnitude conf=0.526Β±0.004 res=0.608
|
| 76 |
+
orbital conf=0.524Β±0.002 res=1.647
|
| 77 |
+
alignment conf=0.485Β±0.014 res=3.110
|
| 78 |
|
| 79 |
+
residual: time=2.03ms norm=0.805 diversity=0.905
|
| 80 |
+
quaternion conf=0.489Β±0.002 res=1.355
|
| 81 |
+
quat_lite conf=0.471Β±0.002 res=1.285
|
| 82 |
+
velocity conf=0.495Β±0.003 res=0.058
|
| 83 |
+
magnitude conf=0.526Β±0.004 res=0.608
|
| 84 |
+
orbital conf=0.524Β±0.002 res=1.647
|
| 85 |
+
alignment conf=0.485Β±0.014 res=3.110
|
| 86 |
|
| 87 |
========================================================================
|
| 88 |
5. GRADIENT HEALTH
|
|
|
|
| 91 |
Flow Loss Grad norm Status
|
| 92 |
ββββββββββββββββββ ββββββββββ ββββββββββββ ββββββββ
|
| 93 |
quaternion mse 8.90e-04 OK
|
| 94 |
+
quat_lite mse 6.98e-04 OK
|
| 95 |
+
velocity mse 3.14e-04 OK
|
| 96 |
+
magnitude mse 8.74e-04 OK
|
| 97 |
+
orbital mse 1.53e-03 OK
|
| 98 |
+
alignment mse 1.55e-02 OK
|
| 99 |
+
quaternion cosine 9.37e-02 OK
|
| 100 |
+
quat_lite cosine 7.39e-02 OK
|
| 101 |
+
velocity cosine 2.66e-02 OK
|
| 102 |
+
magnitude cosine 8.45e-02 OK
|
| 103 |
+
orbital cosine 1.91e-01 OK
|
| 104 |
+
alignment cosine 1.94e+00 OK
|
| 105 |
+
quaternion norm 8.57e-02 OK
|
| 106 |
+
quat_lite norm 6.37e-02 OK
|
| 107 |
+
velocity norm 7.90e-03 OK
|
| 108 |
+
magnitude norm 6.39e-02 OK
|
| 109 |
+
orbital norm 1.45e-01 OK
|
| 110 |
+
alignment norm 2.95e+00 OK
|
| 111 |
|
| 112 |
========================================================================
|
| 113 |
6. ABLATION (100 training steps, rotation target)
|
|
|
|
| 115 |
|
| 116 |
Configuration MSE Params
|
| 117 |
βββββββββββββββββββββββββββββββββββ ββββββββββ ββββββββββ
|
| 118 |
+
QuaternionFlow 0.0011 280,709
|
| 119 |
QuaternionLiteFlow 0.0005 346,245
|
| 120 |
VelocityFlow 0.0061 279,682
|
| 121 |
MagnitudeFlow 0.0081 285,837
|
| 122 |
+
OrbitalFlow 0.0031 91,813
|
| 123 |
+
AlignmentFlow 0.0081 26,010
|
| 124 |
+
Quat + Orbital 0.0009 372,524
|
| 125 |
Velocity + Magnitude 0.0062 565,521
|
| 126 |
+
Orbital + Alignment 0.0026 117,825
|
| 127 |
+
Velocity + Orbital 0.0021 371,497
|
| 128 |
+
Full (weighted) 0.0007 1,310,302
|
| 129 |
+
Full (residual) 0.0007 1,310,302
|
|
|
|
|
|
|
| 130 |
|
| 131 |
========================================================================
|
| 132 |
7. COMPILE COMPATIBILITY
|
|
|
|
| 134 |
|
| 135 |
Flow fullgraph Raw Compiled
|
| 136 |
ββββββββββββββββββββββ ββββββββββββ ββββββββββ ββββββββββββ
|
| 137 |
+
QuaternionFlow OK 274us 200us
|
| 138 |
+
QuaternionLiteFlow OK 199us 181us
|
| 139 |
+
VelocityFlow OK 181us 168us
|
| 140 |
+
MagnitudeFlow OK 328us 351us
|
| 141 |
+
OrbitalFlow OK 376us 347us
|
| 142 |
+
AlignmentFlow OK 348us 362us
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 143 |
|
| 144 |
========================================================================
|
| 145 |
8. MEMORY (B=32, n=128, k=64, d=256)
|
|
|
|
| 151 |
QuaternionLiteFlow 28.3
|
| 152 |
VelocityFlow 48.0
|
| 153 |
MagnitudeFlow 24.2
|
| 154 |
+
OrbitalFlow 10.9
|
| 155 |
+
AlignmentFlow 15.0
|
| 156 |
+
Full ensemble 178.7
|
| 157 |
|
| 158 |
========================================================================
|
| 159 |
Done.
|