Use certificate tools

.github/workflows/python-package-conda.yml

@@ -20,7 +20,8 @@ jobs:
     steps:
     - uses: actions/checkout@v3
       with: 
-        submodules: true
+        token: ${{ secrets.CONSTRAINT_LEARNING_PAT }}
+        submodules: recursive
     - name: Set up Python 3.10
       uses: actions/setup-python@v3
       with:

.gitignore

@@ -12,3 +12,7 @@ venv/
 *.swp
 
 *.code-workspace
+solve_mosek.ptf
+solve_cvxpy*.ptf
+build/
+dist/

.gitmodules

@@ -3,4 +3,7 @@
 	url = [email protected]:utiasASRL/poly_matrix
 [submodule "starloc"]
 	path = starloc
-	url = [email protected]:duembgen/starloc
+	url = [email protected]:utiasASRL/starloc
+[submodule "certifiable-tools"]
+	path = certifiable-tools
+	url = [email protected]:utiasASRL/certifiable-tools

_scripts/generate_all_results.py

@@ -8,12 +8,13 @@
 import matplotlib
 
 try:
-    matplotlib.use("TkAgg")
+    # matplotlib.use("TkAgg")
+    matplotlib.use("Agg")
 except:
     pass
 
 if __name__ == "__main__":
-    n_seeds = 1 # was 10
+    n_seeds = 1  # was 10
     recompute = False
     tightness = True
     scalability = True
@@ -38,12 +39,10 @@
         scalability=scalability,
     )
 
-
-
-    n_successful = 3 # was 100
-    run_datasets_stereo(recompute=False, n_successful=n_successful)
+    n_successful = 3  # was 100
     run_datasets_ro(recompute=False, n_successful=n_successful)
+    run_datasets_stereo(recompute=False, n_successful=n_successful)
 
-    n_successful = 3 # was 10
+    n_successful = 3  # was 10
     run_datasets_ro(recompute=True, n_successful=n_successful)
     run_datasets_stereo(recompute=True, n_successful=n_successful)

_scripts/run_all_study.py

@@ -3,7 +3,7 @@
 import pandas as pd
 import numpy as np
 
-from lifters.learner import Learner
+from auto_template.learner import Learner
 from lifters.mono_lifter import MonoLifter
 from lifters.wahba_lifter import WahbaLifter
 from lifters.stereo2d_lifter import Stereo2DLifter

_scripts/run_datasets_ro.py

@@ -10,12 +10,13 @@
 
 import pandas as pd
 
-from utils.real_experiments import (
+from auto_template.real_experiments import (
     Experiment,
     run_experiments,
     plot_results,
     plot_local_vs_global,
 )
+from auto_template.learner import TOL_RANK_ONE, TOL_REL_GAP
 
 DATASET_ROOT = str(Path(__file__).parent.parent)
 
@@ -74,8 +75,6 @@ def run_all(recompute=RECOMPUTE, n_successful=100, plot_poses=False):
 
         plot_local_vs_global(df, fname_root=fname_root)
 
-        from lifters.learner import TOL_RANK_ONE, TOL_REL_GAP
-
         plot_results(
             df,
             ylabel="SVR",

_scripts/run_datasets_stereo.py

@@ -10,8 +10,13 @@
 except:
     pass
 
-from utils.real_experiments import Experiment
-from utils.real_experiments import run_experiments, plot_results, plot_local_vs_global
+from auto_template.learner import TOL_RANK_ONE, TOL_REL_GAP
+from auto_template.real_experiments import Experiment
+from auto_template.real_experiments import (
+    run_experiments,
+    plot_results,
+    plot_local_vs_global,
+)
 
 DATASET_ROOT = str(Path(__file__).parent.parent)
 MAX_N_LANDMARKS = 8  # 10
@@ -80,8 +85,6 @@ def run_all(recompute=RECOMPUTE, n_successful=10):
         # cost below is found empirically
         plot_local_vs_global(df, fname_root=fname_root, cost_thresh=1e3)
 
-        from lifters.learner import TOL_RANK_ONE, TOL_REL_GAP
-
         plot_results(
             df,
             ylabel="RDG",

_scripts/run_other_study.py

@@ -1,14 +1,14 @@
 import numpy as np
 
-from utils.sim_experiments import plot_scalability, save_table
-from lifters.learner import Learner
-from lifters.mono_lifter import MonoLifter
-from utils.plotting_tools import savefig
-
-from utils.sim_experiments import (
+from auto_template.learner import Learner
+from auto_template.sim_experiments import (
+    plot_scalability,
     run_oneshot_experiment,
     run_scalability_new,
 )
+from lifters.mono_lifter import MonoLifter
+from utils.plotting_tools import savefig
+
 
 DEBUG = False
 

_scripts/run_range_only_study.py

@@ -1,11 +1,11 @@
 import numpy as np
 
-from lifters.learner import Learner
+from auto_template.learner import Learner
 from lifters.range_only_lifters import RangeOnlyLocLifter
 
 from utils.plotting_tools import savefig
-from utils.sim_experiments import plot_scalability, save_table
-from utils.sim_experiments import run_oneshot_experiment, run_scalability_new
+from auto_template.sim_experiments import plot_scalability
+from auto_template.sim_experiments import run_oneshot_experiment, run_scalability_new
 
 
 n_positions = 3
@@ -15,7 +15,7 @@
 DEBUG = False
 
 RESULTS_DIR = "_results"
-#RESULTS_DIR = "_results_server"
+# RESULTS_DIR = "_results_server"
 
 
 def range_only_tightness():
@@ -86,9 +86,7 @@ def range_only_scalability_new(n_seeds, recompute):
         fname_root = f"{RESULTS_DIR}/scalability_{learner.lifter}"
 
         df_sub = df[df.type != "from scratch"]["t solve SDP"]
-        fig, axs = plot_scalability(
-            df, log=True, start="t ", legend_idx=1
-        )
+        fig, axs = plot_scalability(df, log=True, start="t ", legend_idx=1)
 
         # [ax.set_ylim(10, 1000) for ax in axs.values()]
 

_scripts/run_stereo_study.py

@@ -6,23 +6,22 @@
 # matplotlib.use('Agg') # non-interactive
 # plt.ioff()
 
-from utils.sim_experiments import (
+from auto_template.learner import Learner
+from auto_template.sim_experiments import (
+    plot_scalability,
     run_scalability_new,
     run_oneshot_experiment,
     run_scalability_plot,
 )
-from utils.sim_experiments import plot_scalability, save_table
-from utils.plotting_tools import savefig
-
-from lifters.learner import Learner
 from lifters.stereo2d_lifter import Stereo2DLifter
 from lifters.stereo3d_lifter import Stereo3DLifter
 from utils.plotting_tools import savefig
 
 DEBUG = False
 
 RESULTS_DIR = "_results"
-#RESULTS_DIR = "_results_server"
+# RESULTS_DIR = "_results_server"
+
 
 def stereo_tightness(d=2, n_landmarks=None):
     """
@@ -106,7 +105,7 @@ def stereo_scalability_new(n_seeds, recompute, d=2):
         param_list=n_landmarks_list,
         n_seeds=n_seeds,
         recompute=recompute,
-        results_folder=RESULTS_DIR
+        results_folder=RESULTS_DIR,
     )
     if df is None:
         return