imt-ring 1.6.20__tar.gz → 1.6.21__tar.gz

{imt_ring-1.6.20 → imt_ring-1.6.21}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: imt-ring
-Version: 1.6.20
+Version: 1.6.21
 Summary: RING: Recurrent Inertial Graph-based Estimator
 Author-email: Simon Bachhuber <simon.bachhuber@fau.de>
 Project-URL: Homepage, https://github.com/SimiPixel/ring

{imt_ring-1.6.20 → imt_ring-1.6.21}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "imt-ring"
-version = "1.6.20"
+version = "1.6.21"
 authors = [
   { name="Simon Bachhuber", email="simon.bachhuber@fau.de" },
 ]

{imt_ring-1.6.20 → imt_ring-1.6.21}/src/imt_ring.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: imt-ring
-Version: 1.6.20
+Version: 1.6.21
 Summary: RING: Recurrent Inertial Graph-based Estimator
 Author-email: Simon Bachhuber <simon.bachhuber@fau.de>
 Project-URL: Homepage, https://github.com/SimiPixel/ring

{imt_ring-1.6.20 → imt_ring-1.6.21}/src/ring/algorithms/_random.py

@@ -199,7 +199,7 @@ def random_position_over_time(
     POS = jnp.zeros((int(T // t_min) + 1, 2))
     POS = POS.at[0, 1].set(POS_0)
 
-    val_outer = (1, 0.0, 0.0, 0.0, 0.0, key, POS)
+    val_outer = (1, 0.0, 0.0, POS_0, POS_0, key, POS)
     end, *_, consume, POS = jax.lax.while_loop(cond_fn_outer, body_fn_outer, val_outer)
     POS = jnp.where(
         (jnp.arange(len(POS)) < end)[:, None],

{imt_ring-1.6.20 → imt_ring-1.6.21}/src/ring/algorithms/generator/base.py

@@ -34,6 +34,8 @@ class RCMG:
         add_y_relpose: bool = False,
         add_y_rootincl: bool = False,
         add_y_rootincl_kwargs: dict = dict(),
+        add_y_rootfull: bool = False,
+        add_y_rootfull_kwargs: dict = dict(),
         sys_ml: Optional[base.System] = None,
         randomize_positions: bool = False,
         randomize_motion_artifacts: bool = False,
@@ -73,6 +75,8 @@ class RCMG:
                     add_y_relpose=add_y_relpose,
                     add_y_rootincl=add_y_rootincl,
                     add_y_rootincl_kwargs=add_y_rootincl_kwargs,
+                    add_y_rootfull=add_y_rootfull,
+                    add_y_rootfull_kwargs=add_y_rootfull_kwargs,
                     sys_ml=sys_ml,
                     randomize_positions=randomize_positions,
                     randomize_motion_artifacts=randomize_motion_artifacts,
@@ -279,6 +283,8 @@ def _build_mconfig_batched_generator(
     add_y_relpose: bool,
     add_y_rootincl: bool,
     add_y_rootincl_kwargs: dict,
+    add_y_rootfull: bool,
+    add_y_rootfull_kwargs: dict,
     sys_ml: base.System,
     randomize_positions: bool,
     randomize_motion_artifacts: bool,
@@ -365,7 +371,11 @@ def _build_mconfig_batched_generator(
         if add_y_relpose:
             pipe.append(finalize_fns.RelPose(sys_noimu))
         if add_y_rootincl:
+            assert not add_y_rootfull
             pipe.append(finalize_fns.RootIncl(sys_noimu, **add_y_rootincl_kwargs))
+        if add_y_rootfull:
+            assert not add_y_rootincl
+            pipe.append(finalize_fns.RootFull(sys_noimu, **add_y_rootfull_kwargs))
         if use_link_number_in_Xy:
             pipe.append(finalize_fns.Names2Indices(sys_noimu))
 

{imt_ring-1.6.20 → imt_ring-1.6.21}/src/ring/algorithms/generator/finalize_fns.py

@@ -88,6 +88,18 @@ class RootIncl:
         return (X, y), (key, q, x, sys_x)
 
 
+class RootFull:
+    def __init__(self, sys: base.System, **kwargs):
+        self.sys = sys
+        self.kwargs = kwargs
+
+    def __call__(self, Xy, extras):
+        (X, y), (key, q, x, sys_x) = Xy, extras
+        y_root_incl = sensors.root_full(self.sys, x, sys_x, **self.kwargs)
+        y = utils.dict_union(y, y_root_incl)
+        return (X, y), (key, q, x, sys_x)
+
+
 _default_imu_kwargs = dict(
     noisy=True,
     low_pass_filter_pos_f_cutoff=13.5,

{imt_ring-1.6.20 → imt_ring-1.6.21}/src/ring/ml/train.py

@@ -167,7 +167,10 @@ def train_fn(
         tbp=tbp,
     )
 
-    default_callbacks = []
+    # always log, because we also want `i_epsiode` to be logged in wandb
+    default_callbacks = [
+        ml_callbacks.LogEpisodeTrainingLoopCallback(callback_kill_after_episode)
+    ]
     if metrices is not None:
         eval_fn = _build_eval_fn(metrices, filter, link_names)
         default_callbacks.append(_DefaultEvalFnCallback(eval_fn))
@@ -192,11 +195,6 @@ def train_fn(
     if callback_kill_if_nan:
         default_callbacks.append(ml_callbacks.NanKillRunCallback())
 
-    # always log, because we also want `i_epsiode` to be logged in wandb
-    default_callbacks.append(
-        ml_callbacks.LogEpisodeTrainingLoopCallback(callback_kill_after_episode)
-    )
-
     if callback_kill_after_seconds is not None:
         default_callbacks.append(
             ml_callbacks.TimingKillRunCallback(callback_kill_after_seconds)