multipers 2.3.5b1__tar.gz → 2.3.5b2__tar.gz

{multipers-2.3.5b1/multipers.egg-info → multipers-2.3.5b2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: multipers
-Version: 2.3.5b1
+Version: 2.3.5b2
 Summary: Multiparameter Topological Persistence for Machine Learning
 Author-email: David Loiseaux <david.lapous@proton.me>, Hannah Schreiber <hannah.schreiber@inria.fr>
 Maintainer-email: David Loiseaux <david.lapous@proton.me>

{multipers-2.3.5b1 → multipers-2.3.5b2}/multipers/array_api/numpy.py

@@ -22,6 +22,13 @@ unique = _np.unique
 inf = _np.inf
 searchsorted = _np.searchsorted
 LazyTensor = None
+abs = _np.abs
+exp = _np.exp
+
+
+def relu(x):
+    return _np.where(x >= 0, x, 0)
+
 
 # Test keops
 _is_keops_available = None

{multipers-2.3.5b1 → multipers-2.3.5b2}/multipers/array_api/torch.py

@@ -20,6 +20,9 @@ cartesian_product = _t.cartesian_prod
 inf = _t.inf
 searchsorted = _t.searchsorted
 LazyTensor = None
+relu = _t.relu
+abs = _t.abs
+exp = _t.exp
 
 
 _is_keops_available = None

{multipers-2.3.5b1 → multipers-2.3.5b2}/multipers/grids.pyx

@@ -493,3 +493,27 @@ def _push_pts_to_lines(pts, basepoints, directions=None, api=None):
     return out
 
 
+def evaluate_mod_in_grid(mod, grid, box=None):
+    """Given an MMA module, pushes it into the specified grid.
+    Useful for e.g., make it differentiable.
+
+    Input
+    -----
+     - mod: PyModule
+     - grid: Iterable of 1d array, for num_parameters
+    Ouput
+    -----
+    torch-compatible module in the format:
+    (num_degrees) x (num_interval of degree) x ((num_birth, num_parameter), (num_death, num_parameters))
+
+    """
+    (birth_sizes, death_sizes), births, deaths = mod.to_flat_idx(grid)
+    births = evaluate_in_grid(births, grid)
+    deaths = evaluate_in_grid(deaths, grid)
+    diff_mod = tuple(
+        zip(
+            births.split_with_sizes(birth_sizes.tolist()),
+            deaths.split_with_sizes(death_sizes.tolist()),
+        )
+    )
+    return diff_mod

{multipers-2.3.5b1 → multipers-2.3.5b2/multipers.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: multipers
-Version: 2.3.5b1
+Version: 2.3.5b2
 Summary: Multiparameter Topological Persistence for Machine Learning
 Author-email: David Loiseaux <david.lapous@proton.me>, Hannah Schreiber <hannah.schreiber@inria.fr>
 Maintainer-email: David Loiseaux <david.lapous@proton.me>

{multipers-2.3.5b1 → multipers-2.3.5b2}/pyproject.toml

@@ -9,7 +9,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "multipers"
-version = "2.3.5b1"
+version = "2.3.5b2"
 description = "Multiparameter Topological Persistence for Machine Learning"
 readme = "README.md"
 authors = [

{multipers-2.3.5b1 → multipers-2.3.5b2}/multipers/slicer.pyx

@@ -20815,7 +20815,7 @@ def _from_bitmapf64(image, **slicer_kwargs):
             #     print(f"F = {np.asarray(F[i])}")
     slicer = _Slicer(gen_maps, gen_dims, filtration_values)
     return slicer
-def _from_bitmapi32(image, **slicer_kwargs):
+def _from_bitmapi64(image, **slicer_kwargs):
     from multipers import Slicer
     dtype = slicer_kwargs.get("dtype", image.dtype)
     slicer_kwargs["dtype"] = dtype
@@ -20833,9 +20833,9 @@ def _from_bitmapi32(image, **slicer_kwargs):
     cdef cset[unsigned int] vertices
 
     cdef unsigned int num_gens = gen_dims.size()
-    filtration_values = np.zeros(shape=(num_gens, num_parameters), dtype = np.int32) - _Slicer._inf_value()
-    cdef int32_t[:,:] F = filtration_values
-    cdef int32_t[:,:] c_img = image.reshape(-1,num_parameters)
+    filtration_values = np.zeros(shape=(num_gens, num_parameters), dtype = np.int64) - _Slicer._inf_value()
+    cdef int64_t[:,:] F = filtration_values
+    cdef int64_t[:,:] c_img = image.reshape(-1,num_parameters)
     with nogil:
         for i in range(num_gens):
             # with gil:
@@ -20853,7 +20853,7 @@ def _from_bitmapi32(image, **slicer_kwargs):
             #     print(f"F = {np.asarray(F[i])}")
     slicer = _Slicer(gen_maps, gen_dims, filtration_values)
     return slicer
-def _from_bitmapf32(image, **slicer_kwargs):
+def _from_bitmapi32(image, **slicer_kwargs):
     from multipers import Slicer
     dtype = slicer_kwargs.get("dtype", image.dtype)
     slicer_kwargs["dtype"] = dtype
@@ -20871,9 +20871,9 @@ def _from_bitmapf32(image, **slicer_kwargs):
     cdef cset[unsigned int] vertices
 
     cdef unsigned int num_gens = gen_dims.size()
-    filtration_values = np.zeros(shape=(num_gens, num_parameters), dtype = np.float32) - _Slicer._inf_value()
-    cdef float[:,:] F = filtration_values
-    cdef float[:,:] c_img = image.reshape(-1,num_parameters)
+    filtration_values = np.zeros(shape=(num_gens, num_parameters), dtype = np.int32) - _Slicer._inf_value()
+    cdef int32_t[:,:] F = filtration_values
+    cdef int32_t[:,:] c_img = image.reshape(-1,num_parameters)
     with nogil:
         for i in range(num_gens):
             # with gil:
@@ -20891,7 +20891,7 @@ def _from_bitmapf32(image, **slicer_kwargs):
             #     print(f"F = {np.asarray(F[i])}")
     slicer = _Slicer(gen_maps, gen_dims, filtration_values)
     return slicer
-def _from_bitmapi64(image, **slicer_kwargs):
+def _from_bitmapf32(image, **slicer_kwargs):
     from multipers import Slicer
     dtype = slicer_kwargs.get("dtype", image.dtype)
     slicer_kwargs["dtype"] = dtype
@@ -20909,9 +20909,9 @@ def _from_bitmapi64(image, **slicer_kwargs):
     cdef cset[unsigned int] vertices
 
     cdef unsigned int num_gens = gen_dims.size()
-    filtration_values = np.zeros(shape=(num_gens, num_parameters), dtype = np.int64) - _Slicer._inf_value()
-    cdef int64_t[:,:] F = filtration_values
-    cdef int64_t[:,:] c_img = image.reshape(-1,num_parameters)
+    filtration_values = np.zeros(shape=(num_gens, num_parameters), dtype = np.float32) - _Slicer._inf_value()
+    cdef float[:,:] F = filtration_values
+    cdef float[:,:] c_img = image.reshape(-1,num_parameters)
     with nogil:
         for i in range(num_gens):
             # with gil:
@@ -20934,12 +20934,12 @@ def from_bitmap(img, **kwargs):
     img = np.asarray(img)
     if img.dtype == np.float64:
         return _from_bitmapf64(img, **kwargs)
+    if img.dtype == np.int64:
+        return _from_bitmapi64(img, **kwargs)
     if img.dtype == np.int32:
         return _from_bitmapi32(img, **kwargs)
     if img.dtype == np.float32:
         return _from_bitmapf32(img, **kwargs)
-    if img.dtype == np.int64:
-        return _from_bitmapi64(img, **kwargs)
     raise ValueError(f"Invalid dtype. Got {img.dtype=}, was expecting {available_dtype=}.")
 
 def from_function_delaunay(