Fixed 4PL and 5PL model Jacobians, simplified dropna, and improved Ca…

…lCurve.

build/lib/waltlabtools/__init__.py

@@ -0,0 +1,5 @@
+from . import mosaic
+from .cal_curve import *
+from .core import *
+from .model import *
+from .read import *

build/lib/waltlabtools/_backend.py

@@ -0,0 +1,33 @@
+import sys
+from collections.abc import Callable
+
+_MODULES_CHECK: dict[str, bool] = {
+    "jax": "jax" in sys.modules,
+    "numba": True,
+}
+
+if _MODULES_CHECK["jax"]:
+    import jax.numpy as np
+    from jax import jit
+    from jax.scipy.special import gammaln
+
+    def gmean(a, axis: int = 0, dtype=None, weights=None):
+        """Geometric mean."""
+        return np.exp(np.average(np.log(a), axis=axis, weights=weights))
+
+    import warnings
+
+    warnings.warn("JAX backend is experimental and may not work as expected.")  # type: ignore
+else:
+    import numpy as np
+    from scipy.special import gammaln
+    from scipy.stats import gmean
+
+    try:
+        from numba import jit  # type: ignore
+    except ImportError:
+        _MODULES_CHECK["numba"] = False
+
+        def jit(fun: Callable) -> Callable:
+            """Dummy jit decorator for when numba/jax not installed."""
+            return fun