replace hard-coded const multipliers for NLDFs with new normalizers

mir-group · Jun 21, 2024 · 26e01e7 · 26e01e7
1 parent 7cbeb4b
commit 26e01e7
Show file tree

Hide file tree

Showing 4 changed files with 13 additions and 17 deletions.
diff --git a/ciderpress/gpaw/analysis.py b/ciderpress/gpaw/analysis.py
@@ -580,6 +580,7 @@ def calculate_6d_integral_deriv(
         if n_g is not None:
             dfeatdf = np.zeros([nexp - 1] + list(n_g.shape))
         for i in range(nexp - 1):
+            # TODO this should be changed to use the feature normalizers
             const = ((self.consts[i, 1] + self.consts[-1, 1]) / 2) ** 1.5
             const = const + self.consts[i, 0] / (self.consts[i, 0] + const)
             for ind, a in enumerate(self.alphas):

diff --git a/ciderpress/gpaw/atom_analysis_utils.py b/ciderpress/gpaw/atom_analysis_utils.py
@@ -389,6 +389,7 @@ def _get_paw_helper2(
         dydf_obg = dFdf_oag
     Nalpha = self.Nalpha
 
+    # TODO consts should be removed to make use of feat normalizers
     x_sig = np.zeros((nspin, nfeat, ngrid))
     dxdf_oig = np.zeros((norb, nfeat, ngrid))
     for s in range(nspin):

diff --git a/ciderpress/gpaw/cider_fft.py b/ciderpress/gpaw/cider_fft.py
@@ -261,17 +261,15 @@ def call_xc_kernel(
             X0T[:, start : start + nfeat_tmp] = nspin * feat_sg
             start += nfeat_tmp
 
-        X0TN = (
-            X0T  # self.mlfunc.settings.normalizers.get_normalized_feature_vector(X0T)
-        )
+        X0TN = self.mlfunc.settings.normalizers.get_normalized_feature_vector(X0T)
         exc_ml, dexcdX0TN_ml = self.mlfunc(X0TN, rhocut=self.rhocut)
         xmix = self.xmix  # / rho.shape[0]
         exc_ml *= xmix
         dexcdX0TN_ml *= xmix
-        # vxc_ml = self.mlfunc.settings.normalizers.get_derivative_wrt_unnormed_features(
-        #     X0T, dexcdX0TN_ml
-        # )
-        vxc_ml = dexcdX0TN_ml
+        vxc_ml = self.mlfunc.settings.normalizers.get_derivative_wrt_unnormed_features(
+            X0T, dexcdX0TN_ml
+        )
+        # vxc_ml = dexcdX0TN_ml
         e_g[:] += exc_ml
 
         start = 0
@@ -975,8 +973,6 @@ def calculate_6d_integral_fwd(self, n_g, cider_exp, c_abi=None):
             else:
                 i_g = None
                 dq0_g = None
-            const = ((self.consts[i, 1] + self.consts[-1, 1]) / 2) ** 1.5
-            const = const + self.consts[i, 0] / (self.consts[i, 0] + const)
             for ind, a in enumerate(self.alphas):
                 self.timer.start("COEFS")
                 pa_g, dpa_g = fnlxc.eval_cubic_interp(
@@ -987,8 +983,8 @@ def calculate_6d_integral_fwd(self, n_g, cider_exp, c_abi=None):
                 self.timer.stop()
                 p_iag[i, a] = pa_g
 
-                feat[i, :] += const * pa_g * self.rbuf_ag[a]
-                dfeat[i, :] += const * dpa_g * self.rbuf_ag[a]
+                feat[i, :] += pa_g * self.rbuf_ag[a]
+                dfeat[i, :] += dpa_g * self.rbuf_ag[a]
 
         self.timer.start("6d comm fwd")
         if n_g is not None:
@@ -1041,11 +1037,9 @@ def calculate_6d_integral_bwd(
             for a in self.alphas:
                 self.rbuf_ag[a][:] = 0.0
         for i in range(nexp - 1):
-            const = ((self.consts[i, 1] + self.consts[-1, 1]) / 2) ** 1.5
-            const = const + self.consts[i, 0] / (self.consts[i, 0] + const)
             vfeati_g = self.domain_world2cider(vfeat_g[i])
             for a in alphas:
-                self.rbuf_ag[a][:] += vfeati_g * const * p_iag[i, a]
+                self.rbuf_ag[a][:] += vfeati_g * p_iag[i, a]
         self.timer.stop()
         if compute_stress:
             self.calculate_6d_stress_integral()

diff --git a/ciderpress/gpaw/paw_cider_kernel.py b/ciderpress/gpaw/paw_cider_kernel.py
@@ -173,9 +173,9 @@ def get_paw_atom_contribs_en(self, n_sg, sigma_xg, y_sbg, F_sag, ae=True):
                     x_sig[s, i] += pa_g * y_sbg[s, a]
                     xd_sig[s, i] += dpa_g * y_sbg[s, a]
                     p_siag[s, i, a] = pa_g
-                x_sig[s, i] *= ((self.consts[i, 1] + self.consts[-1, 1]) / 2) ** 1.5
-                xd_sig[s, i] *= ((self.consts[i, 1] + self.consts[-1, 1]) / 2) ** 1.5
-                p_siag[s, i] *= ((self.consts[i, 1] + self.consts[-1, 1]) / 2) ** 1.5
+                # x_sig[s, i] *= ((self.consts[i, 1] + self.consts[-1, 1]) / 2) ** 1.5
+                # xd_sig[s, i] *= ((self.consts[i, 1] + self.consts[-1, 1]) / 2) ** 1.5
+                # p_siag[s, i] *= ((self.consts[i, 1] + self.consts[-1, 1]) / 2) ** 1.5
                 dgdn_sig[s, i] = xd_sig[s, i] * dadn
                 dgdsigma_sig[s, i] = xd_sig[s, i] * dadsigma