Merge branch 'main' into add_tests

torchcfm/conditional_flow_matching.py

@@ -7,6 +7,7 @@
 
 import math
 import warnings
+from typing import Union
 
 import torch
 
@@ -48,12 +49,12 @@ class ConditionalFlowMatcher:
     - score function $\nabla log p_t(x|x0, x1)$
     """
 
-    def __init__(self, sigma: float = 0.0):
+    def __init__(self, sigma: Union[float, int] = 0.0):
         r"""Initialize the ConditionalFlowMatcher class. It requires the hyper-parameter $\sigma$.
 
         Parameters
         ----------
-        sigma : float
+        sigma : Union[float, int]
         """
         self.sigma = sigma
 
@@ -216,15 +217,15 @@ class ExactOptimalTransportConditionalFlowMatcher(ConditionalFlowMatcher):
     It overrides the sample_location_and_conditional_flow.
     """
 
-    def __init__(self, sigma: float = 0.0):
+    def __init__(self, sigma: Union[float, int] = 0.0):
         r"""Initialize the ConditionalFlowMatcher class. It requires the hyper-parameter $\sigma$.
 
         Parameters
         ----------
-        sigma : float
+        sigma : Union[float, int]
         ot_sampler: exact OT method to draw couplings (x0, x1) (see Eq.(17) [1]).
         """
-        self.sigma = sigma
+        super().__init__(sigma)
         self.ot_sampler = OTPlanSampler(method="exact")
 
     def sample_location_and_conditional_flow(self, x0, x1, return_noise=False):
@@ -383,13 +384,13 @@ class SchrodingerBridgeConditionalFlowMatcher(ConditionalFlowMatcher):
     sample_location_and_conditional_flow functions.
     """
 
-    def __init__(self, sigma: float = 1.0, ot_method="exact"):
+    def __init__(self, sigma: Union[float, int] = 1.0, ot_method="exact"):
         r"""Initialize the SchrodingerBridgeConditionalFlowMatcher class. It requires the hyper-
         parameter $\sigma$ and the entropic OT map.
 
         Parameters
         ----------
-        sigma : float
+        sigma : Union[float, int]
         ot_sampler: exact OT method to draw couplings (x0, x1) (see Eq.(17) [1]).
             we use exact as the default as we found this to perform better
             (more accurate and faster) in practice for reasonable batch sizes.
@@ -400,7 +401,7 @@ def __init__(self, sigma: float = 1.0, ot_method="exact"):
             raise ValueError(f"Sigma must be strictly positive, got {sigma}.")
         elif sigma < 1e-3:
             warnings.warn("Small sigma values may lead to numerical instability.")
-        self.sigma = sigma
+        super().__init__(sigma)
         self.ot_method = ot_method
         self.ot_sampler = OTPlanSampler(method=ot_method, reg=2 * self.sigma**2)