classy-szfast 0.0.25.post2__py3-none-any.whl → 0.0.25.post4__py3-none-any.whl

classy_szfast/classy_szfast.py

@@ -1,7 +1,7 @@
 from .utils import *
 from .config import *
 import numpy as np
-from .emulators_meta_data import emulator_dict, dofftlog_alphas, cp_l_max_scalars
+from .emulators_meta_data import emulator_dict, dofftlog_alphas, cp_l_max_scalars, cosmo_model_list
 from .cosmopower import cp_tt_nn, cp_te_nn, cp_ee_nn, cp_ee_nn, cp_pp_nn, cp_pknl_nn, cp_pkl_nn, cp_der_nn, cp_da_nn, cp_h_nn, cp_s8_nn, cp_pkl_fftlog_alphas_real_nn, cp_pkl_fftlog_alphas_imag_nn, cp_pkl_fftlog_alphas_nus
 from .cosmopower_jax import cp_tt_nn_jax, cp_te_nn_jax, cp_ee_nn_jax, cp_ee_nn_jax, cp_pp_nn_jax, cp_pknl_nn_jax, cp_pkl_nn_jax, cp_der_nn_jax, cp_da_nn_jax, cp_h_nn_jax, cp_s8_nn_jax
 from .pks_and_sigmas import *
@@ -76,13 +76,10 @@ class Class_szfast(object):
 
         self.jax_mode = params_settings["jax"]
 
+ 
         # print(f"JAX mode: {self.jax_mode}")
         
 
-        
-
-
-
         # cosmopower emulators
         # self.cp_path_to_cosmopower_organization = path_to_cosmopower_organization + '/'
         self.cp_tt_nn = cp_tt_nn
@@ -90,14 +87,70 @@ class Class_szfast(object):
         self.cp_ee_nn = cp_ee_nn
         self.cp_pp_nn = cp_pp_nn
         self.cp_pknl_nn  = cp_pknl_nn
-        self.cp_pkl_nn = cp_pkl_nn
         self.cp_der_nn = cp_der_nn
-        self.cp_da_nn = cp_da_nn
+       
+        self.cp_lmax = cp_l_max_scalars
+
+        cosmo_model_dict = {i: model for i, model in enumerate(cosmo_model_list)}
+
+        if (cosmo_model_dict[params_settings['cosmo_model']] == 'ede-v2'):
+        
+            self.cszfast_pk_grid_zmax = 20.
+            self.cszfast_pk_grid_kmin = 5e-4
+            self.cszfast_pk_grid_kmax = 10. 
+            self.cp_kmax = self.cszfast_pk_grid_kmax
+            self.cp_kmin = self.cszfast_pk_grid_kmin
+            # self.logger.info(f">>> using kmin = {self.cp_kmin}")
+            # self.logger.info(f">>> using kmax = {self.cp_kmax}")
+            # self.logger.info(f">>> using zmax = {self.cszfast_pk_grid_zmax}")
+        
+        else:
+        
+            self.cszfast_pk_grid_zmax = 5. #  max z of our pk emulators (sept 23)
+            self.cszfast_pk_grid_kmin = 1e-4
+            self.cszfast_pk_grid_kmax = 50.
+            self.cp_kmax = self.cszfast_pk_grid_kmax
+            self.cp_kmin = self.cszfast_pk_grid_kmin
+            # self.logger.info(f">>> using kmin = {self.cp_kmin}")
+            # self.logger.info(f">>> using kmax = {self.cp_kmax}")
+            # self.logger.info(f">>> using zmax = {self.cszfast_pk_grid_zmax}")
         
         if self.jax_mode:
             self.cp_h_nn = cp_h_nn_jax
+            self.cp_da_nn = cp_da_nn_jax
+            self.cp_pkl_nn = cp_pkl_nn_jax
+
+            self.pi = jnp.pi
+            self.transpose = jnp.transpose
+            self.asarray = jnp.asarray
+
+            self.linspace = jnp.linspace
+            self.geomspace = jnp.geomspace
+            self.arange = jnp.arange
+            self.zeros = jnp.zeros
+
+
         else:
             self.cp_h_nn = cp_h_nn
+            self.cp_da_nn = cp_da_nn
+            self.cp_pkl_nn = cp_pkl_nn
+
+            self.pi = np.pi
+            self.transpose = np.transpose
+
+            self.linspace = np.linspace
+            self.geomspace = np.geomspace
+            self.arange = np.arange
+            self.zeros = np.zeros
+            self.asarray = np.asarray
+
+
+
+        self.cp_ls = self.arange(2,self.cp_lmax+1)
+        self.cp_predicted_tt_spectrum =self.zeros(self.cp_lmax)
+        self.cp_predicted_te_spectrum =self.zeros(self.cp_lmax)
+        self.cp_predicted_ee_spectrum =self.zeros(self.cp_lmax)
+        self.cp_predicted_pp_spectrum =self.zeros(self.cp_lmax)
 
         self.cp_s8_nn = cp_s8_nn
 
@@ -112,20 +165,6 @@ class Class_szfast(object):
         self.use_Amod = 0
         self.Amod = 0 
 
-        self.cp_lmax = cp_l_max_scalars
-        self.cp_ls = np.arange(2,self.cp_lmax+1)
-
-
-
-        
-        cosmo_model_dict = {0: 'lcdm',
-                            1: 'mnu',
-                            2: 'neff',
-                            3: 'wcdm',
-                            4: 'ede',
-                            5: 'mnu-3states',
-                            6: 'ede-v2'
-                            }
         
 
         if cosmo_model_dict[params_settings['cosmo_model']] == 'ede-v2':
@@ -138,60 +177,31 @@ class Class_szfast(object):
             self.cp_ndspl_k = 10
             self.cp_nk = 5000
 
-        self.cp_predicted_tt_spectrum =np.zeros(self.cp_lmax)
-        self.cp_predicted_te_spectrum =np.zeros(self.cp_lmax)
-        self.cp_predicted_ee_spectrum =np.zeros(self.cp_lmax)
-        self.cp_predicted_pp_spectrum =np.zeros(self.cp_lmax)
-
 
         self.cszfast_ldim = 20000 # used for the cls arrays
-
         self.cszfast_pk_grid_nz = 100 # has to be same as narraySZ, i.e., ndim_redshifts; it is setup hereafter if ndim_redshifts is passed
         
 
-
-        if (cosmo_model_dict[params_settings['cosmo_model']] == 'ede-v2'):
-        
-            self.cszfast_pk_grid_zmax = 20.
-            self.cszfast_pk_grid_kmin = 5e-4
-            self.cszfast_pk_grid_kmax = 10. 
-            self.cp_kmax = self.cszfast_pk_grid_kmax
-            self.cp_kmin = self.cszfast_pk_grid_kmin
-            # self.logger.info(f">>> using kmin = {self.cp_kmin}")
-            # self.logger.info(f">>> using kmax = {self.cp_kmax}")
-            # self.logger.info(f">>> using zmax = {self.cszfast_pk_grid_zmax}")
-        
-        else:
-        
-            self.cszfast_pk_grid_zmax = 5. #  max z of our pk emulators (sept 23)
-            self.cszfast_pk_grid_kmin = 1e-4
-            self.cszfast_pk_grid_kmax = 50.
-            self.cp_kmax = self.cszfast_pk_grid_kmax
-            self.cp_kmin = self.cszfast_pk_grid_kmin
-            # self.logger.info(f">>> using kmin = {self.cp_kmin}")
-            # self.logger.info(f">>> using kmax = {self.cp_kmax}")
-            # self.logger.info(f">>> using zmax = {self.cszfast_pk_grid_zmax}")
-
-        self.cszfast_pk_grid_z = np.linspace(0.,self.cszfast_pk_grid_zmax,self.cszfast_pk_grid_nz)
+        self.cszfast_pk_grid_z = self.linspace(0.,self.cszfast_pk_grid_zmax,self.cszfast_pk_grid_nz)
         self.cszfast_pk_grid_ln1pz = np.log(1.+self.cszfast_pk_grid_z)
 
 
-        self.cszfast_pk_grid_k = np.geomspace(self.cp_kmin,self.cp_kmax,self.cp_nk)[::self.cp_ndspl_k]
+        self.cszfast_pk_grid_k = self.geomspace(self.cp_kmin,self.cp_kmax,self.cp_nk)[::self.cp_ndspl_k]
         
         self.cszfast_pk_grid_lnk = np.log(self.cszfast_pk_grid_k)
         
-        self.cszfast_pk_grid_nk = len(np.geomspace(self.cp_kmin,self.cp_kmax,self.cp_nk)[::self.cp_ndspl_k]) # has to be same as ndimSZ, and the same as dimension of cosmopower pk emulators
+        self.cszfast_pk_grid_nk = len(self.geomspace(self.cp_kmin,self.cp_kmax,self.cp_nk)[::self.cp_ndspl_k]) # has to be same as ndimSZ, and the same as dimension of cosmopower pk emulators
         
         for k,v in params_settings.items():
 
             if k == 'ndim_redshifts':
                 
                 self.cszfast_pk_grid_nz = v
-                self.cszfast_pk_grid_z = np.linspace(0.,self.cszfast_pk_grid_zmax,self.cszfast_pk_grid_nz)
+                self.cszfast_pk_grid_z = self.linspace(0.,self.cszfast_pk_grid_zmax,self.cszfast_pk_grid_nz)
                 self.cszfast_pk_grid_ln1pz = np.log(1.+self.cszfast_pk_grid_z)
 
-                self.cszfast_pk_grid_pknl_flat = np.zeros(self.cszfast_pk_grid_nz*self.cszfast_pk_grid_nk)
-                self.cszfast_pk_grid_pkl_flat = np.zeros(self.cszfast_pk_grid_nz*self.cszfast_pk_grid_nk)
+                self.cszfast_pk_grid_pknl_flat = self.zeros(self.cszfast_pk_grid_nz*self.cszfast_pk_grid_nk)
+                self.cszfast_pk_grid_pkl_flat = self.zeros(self.cszfast_pk_grid_nz*self.cszfast_pk_grid_nk)
 
             if k == 'cosmo_model':
 
@@ -210,13 +220,12 @@ class Class_szfast(object):
         
         else:
 
-            ls = np.arange(2,self.cp_nk+2)[::self.cp_ndspl_k] # jan 10 ndspl
-            dls = ls*(ls+1.)/2./np.pi
+            ls = self.arange(2,self.cp_nk+2)[::self.cp_ndspl_k] # jan 10 ndspl
+            dls = ls*(ls+1.)/2./self.pi
             self.pk_power_fac= (dls)**-1
 
 
-        self.cp_z_interp = np.linspace(0.,20.,5000)
-        self.cp_z_interp_jax = jnp.linspace(0.,20.,5000)
+        self.cp_z_interp = self.linspace(0.,20.,5000)
 
         self.csz_base = None
 
@@ -224,7 +233,7 @@ class Class_szfast(object):
         self.cszfast_zgrid_zmin = 0.
         self.cszfast_zgrid_zmax = 4.
         self.cszfast_zgrid_nz = 250
-        self.cszfast_zgrid = np.linspace(self.cszfast_zgrid_zmin,
+        self.cszfast_zgrid = self.linspace(self.cszfast_zgrid_zmin,
                                          self.cszfast_zgrid_zmax,
                                          self.cszfast_zgrid_nz)
 
@@ -232,14 +241,14 @@ class Class_szfast(object):
         self.cszfast_mgrid_mmin = 1e10
         self.cszfast_mgrid_mmax = 1e15
         self.cszfast_mgrid_nm = 50
-        self.cszfast_mgrid = np.geomspace(self.cszfast_mgrid_mmin,
+        self.cszfast_mgrid = self.geomspace(self.cszfast_mgrid_mmin,
                                           self.cszfast_mgrid_mmax,
                                           self.cszfast_mgrid_nm)
 
         self.cszfast_gas_pressure_xgrid_xmin = 1e-2
         self.cszfast_gas_pressure_xgrid_xmax = 1e2
         self.cszfast_gas_pressure_xgrid_nx = 100
-        self.cszfast_gas_pressure_xgrid = np.geomspace(self.cszfast_gas_pressure_xgrid_xmin,
+        self.cszfast_gas_pressure_xgrid = self.geomspace(self.cszfast_gas_pressure_xgrid_xmin,
                                                        self.cszfast_gas_pressure_xgrid_xmax,
                                                        self.cszfast_gas_pressure_xgrid_nx)
         
@@ -310,7 +319,7 @@ class Class_szfast(object):
         creal = predicted_testing_alphas_creal
         cimag = predicted_testing_alphas_cimag
         Nmax = len(self.cszfast_pk_grid_k)
-        cnew = np.zeros(Nmax+1,dtype=complex)
+        cnew = self.zeros(Nmax+1,dtype=complex)
         for i in range(Nmax+1):
             if i<int(Nmax/2):
                 cnew[i] = complex(creal[i],cimag[i])
@@ -360,20 +369,20 @@ class Class_szfast(object):
 
         nl = len(self.cp_predicted_tt_spectrum)
         cls = {}
-        cls['ell'] = np.arange(20000)
-        cls['tt'] = np.zeros(20000)
-        cls['te'] = np.zeros(20000)
-        cls['ee'] = np.zeros(20000)
-        cls['pp'] = np.zeros(20000)
-        cls['bb'] = np.zeros(20000)
-        lcp = np.asarray(cls['ell'][2:nl+2])
+        cls['ell'] = self.arange(20000)
+        cls['tt'] = self.zeros(20000)
+        cls['te'] = self.zeros(20000)
+        cls['ee'] = self.zeros(20000)
+        cls['pp'] = self.zeros(20000)
+        cls['bb'] = self.zeros(20000)
+        lcp = self.asarray(cls['ell'][2:nl+2])
 
         # print('cosmo_model:',self.cosmo_model,nl)
         if self.cosmo_model == 'ede-v2':
             factor_ttteee = 1./lcp**2 
             factor_pp = 1./lcp**3
         else:
-            factor_ttteee = 1./(lcp*(lcp+1.)/2./np.pi)
+            factor_ttteee = 1./(lcp*(lcp+1.)/2./self.pi)
             factor_pp = 1./(lcp*(lcp+1.))**2.        
 
         self.cp_predicted_tt_spectrum *= factor_ttteee
@@ -400,7 +409,7 @@ class Class_szfast(object):
             cmb_cls_loaded = pickle.load(handle)
         nl_cls_file = len(cmb_cls_loaded['ell'])
         cls_ls = cmb_cls_loaded['ell']
-        dlfac = cls_ls*(cls_ls+1.)/2./np.pi
+        dlfac = cls_ls*(cls_ls+1.)/2./self.pi
         cls_tt = cmb_cls_loaded['tt']*dlfac
         cls_te = cmb_cls_loaded['te']*dlfac
         cls_ee = cmb_cls_loaded['ee']*dlfac
@@ -419,8 +428,6 @@ class Class_szfast(object):
                       **params_values_dict):
  
         z_arr = self.cszfast_pk_grid_z
-
-
         k_arr = self.cszfast_pk_grid_k 
 
         # print(">>> z_arr:",z_arr)
@@ -462,26 +469,32 @@ class Class_szfast(object):
             
                 params_dict_pp = params_dict.copy()
                 params_dict_pp['z_pk_save_nonclass'] = [zp]
-                predicted_pk_spectrum_z.append(self.cp_pkl_nn[self.cosmo_model].predictions_np(params_dict_pp)[0])
+                if self.jax_mode:
+                    predicted_pk_spectrum_z.append(self.cp_pkl_nn[self.cosmo_model].predict(params_dict_pp))
+                else:
+                    predicted_pk_spectrum_z.append(self.cp_pkl_nn[self.cosmo_model].predictions_np(params_dict_pp)[0])
 
                 # if abs(zp-0.5) < 0.01:
                 #   print(">>> predicted_pk_spectrum_z:",predicted_pk_spectrum_z[-1])
                 #   import pprint
                 #   pprint.pprint(params_dict_pp)
        
-        predicted_pk_spectrum = np.asarray(predicted_pk_spectrum_z)
+        predicted_pk_spectrum = self.asarray(predicted_pk_spectrum_z)
 
 
         pk = 10.**predicted_pk_spectrum
 
         pk_re = pk*self.pk_power_fac
-        pk_re = np.transpose(pk_re)
+        pk_re = self.transpose(pk_re)
 
         # print(">>> pk_re:",pk_re)
         # import sys
         # sys.exit(0)
 
-        self.pkl_interp = PowerSpectrumInterpolator(z_arr,k_arr,np.log(pk_re).T,logP=True)
+        if self.jax_mode:
+            self.pkl_interp = None
+        else:
+            self.pkl_interp = PowerSpectrumInterpolator(z_arr,k_arr,np.log(pk_re).T,logP=True)
 
         self.cszfast_pk_grid_pk = pk_re
         self.cszfast_pk_grid_pkl_flat = pk_re.flatten()
@@ -568,7 +581,7 @@ class Class_szfast(object):
         s8z  = self.cp_s8_nn[self.cosmo_model].predictions_np(params_dict)
         # print(self.s8z)
         self.s8z_interp = scipy.interpolate.interp1d(
-                                                    np.linspace(0.,20.,5000),
+                                                    self.linspace(0.,20.,5000),
                                                     s8z[0],
                                                     kind='linear',
                                                     axis=-1,
@@ -607,13 +620,13 @@ class Class_szfast(object):
             params_dict_pp['z_pk_save_nonclass'] = [zp]
             predicted_pk_spectrum_z.append(self.cp_pknl_nn[self.cosmo_model].predictions_np(params_dict_pp)[0])
 
-        predicted_pk_spectrum = np.asarray(predicted_pk_spectrum_z)
+        predicted_pk_spectrum = self.asarray(predicted_pk_spectrum_z)
 
 
         pk = 10.**predicted_pk_spectrum
 
         pk_re = pk*self.pk_power_fac
-        pk_re = np.transpose(pk_re)
+        pk_re = self.transpose(pk_re)
 
 
         self.pknl_interp = PowerSpectrumInterpolator(z_arr,k_arr,np.log(pk_re).T,logP=True)
@@ -629,16 +642,12 @@ class Class_szfast(object):
                            z_asked,
                           params_values_dict=None):
 
-        z_arr = self.cszfast_pk_grid_z
 
         k_arr = self.cszfast_pk_grid_k 
 
         if params_values_dict:
-
             params_values = params_values_dict.copy()
-
         else:
-
             params_values = self.params_for_emulators
 
         update_params_with_defaults(params_values, self.emulator_dict[self.cosmo_model]['default'])
@@ -655,19 +664,21 @@ class Class_szfast(object):
 
         predicted_pk_spectrum_z = []
 
-        z_asked = z_asked
         params_dict_pp = params_dict.copy()
         update_params_with_defaults(params_dict_pp, self.emulator_dict[self.cosmo_model]['default'])
 
         params_dict_pp['z_pk_save_nonclass'] = [z_asked]
-        predicted_pk_spectrum_z.append(self.cp_pkl_nn[self.cosmo_model].predictions_np(params_dict_pp)[0])
+        if self.jax_mode:
+            predicted_pk_spectrum_z.append(self.cp_pkl_nn[self.cosmo_model].predict(params_dict_pp))
+        else:
+            predicted_pk_spectrum_z.append(self.cp_pkl_nn[self.cosmo_model].predictions_np(params_dict_pp)[0])
 
-        predicted_pk_spectrum = np.asarray(predicted_pk_spectrum_z)
+        predicted_pk_spectrum = self.asarray(predicted_pk_spectrum_z)
 
 
         pk = 10.**predicted_pk_spectrum
         pk_re = pk*self.pk_power_fac
-        pk_re = np.transpose(pk_re)
+        pk_re = self.transpose(pk_re)
 
         
         return pk_re, k_arr
@@ -710,12 +721,12 @@ class Class_szfast(object):
         params_dict_pp['z_pk_save_nonclass'] = [z_asked]
         predicted_pk_spectrum_z.append(self.cp_pknl_nn[self.cosmo_model].predictions_np(params_dict_pp)[0])
 
-        predicted_pk_spectrum = np.asarray(predicted_pk_spectrum_z)
+        predicted_pk_spectrum = self.asarray(predicted_pk_spectrum_z)
 
 
         pk = 10.**predicted_pk_spectrum
         pk_re = pk*self.pk_power_fac
-        pk_re = np.transpose(pk_re)
+        pk_re = self.transpose(pk_re)
 
         
         return pk_re, k_arr
@@ -746,19 +757,10 @@ class Class_szfast(object):
             # print("self.cp_predicted_hubble type:", type(self.cp_predicted_hubble))
             # print("self.cp_predicted_hubble",self.cp_predicted_hubble)
 
-            # self.hz_interp = jscipy.interpolate.interp1d(
-            #                                     self.cp_z_interp_jax,
-            #                                     self.cp_predicted_hubble,
-            #                                     kind='linear',
-            #                                     axis=-1,
-            #                                     copy=True,
-            #                                     bounds_error=None,
-            #                                     fill_value=np.nan,
-            #                                     assume_sorted=False)
             
             # Assuming `cp_z_interp` and `cp_predicted_hubble` are JAX arrays
             def hz_interp(x):
-                return jnp.interp(x, self.cp_z_interp_jax, self.cp_predicted_hubble, left=jnp.nan, right=jnp.nan)
+                return jnp.interp(x, self.cp_z_interp, self.cp_predicted_hubble, left=jnp.nan, right=jnp.nan)
 
             self.hz_interp = hz_interp
             # exit()
@@ -785,45 +787,60 @@ class Class_szfast(object):
         update_params_with_defaults(params_values, self.emulator_dict[self.cosmo_model]['default'])
 
         params_dict = {}
-
         for k,v in zip(params_values.keys(),params_values.values()):
-        
             params_dict[k]=[v]
 
         if 'm_ncdm' in params_dict.keys():
             if isinstance(params_dict['m_ncdm'][0],str): 
                 params_dict['m_ncdm'] =  [float(params_dict['m_ncdm'][0].split(',')[0])]
 
-        # deal with different scaling of DA in different model from emulator training
-        if self.cosmo_model == 'ede-v2':
+        if self.jax_mode: 
+            # print("JAX MODE in chi")
+            # self.cp_da_nn[self.cosmo_model].log = False
+            self.cp_predicted_da = self.cp_da_nn[self.cosmo_model].predict(params_dict)
+            # print("self.cp_predicted_da",self.cp_predicted_da)
+
+            if self.cosmo_model == 'ede-v2':
+                # print('ede-v2 case')
+                self.cp_predicted_da = jnp.insert(self.cp_predicted_da, 0, 0)
+            self.cp_predicted_da *= (1.+self.cp_z_interp)
+    
+            def chi_interp(x):
+                return jnp.interp(x, self.cp_z_interp, self.cp_predicted_da, left=jnp.nan, right=jnp.nan)
+
+            self.chi_interp = chi_interp
 
-            self.cp_predicted_da  = self.cp_da_nn[self.cosmo_model].ten_to_predictions_np(params_dict)[0]
-            self.cp_predicted_da = np.insert(self.cp_predicted_da, 0, 0)
-        
         else:
-        
-            self.cp_predicted_da  = self.cp_da_nn[self.cosmo_model].predictions_np(params_dict)[0]
-        
+            # deal with different scaling of DA in different model from emulator training
+            if self.cosmo_model == 'ede-v2':
 
-        self.chi_interp = scipy.interpolate.interp1d(
-                                                    self.cp_z_interp,
-                                                    self.cp_predicted_da*(1.+self.cp_z_interp),
-                                                    kind='linear',
-                                                    axis=-1,
-                                                    copy=True,
-                                                    bounds_error=None,
-                                                    fill_value=np.nan,
-                                                    assume_sorted=False)
+                self.cp_predicted_da  = self.cp_da_nn[self.cosmo_model].ten_to_predictions_np(params_dict)[0]
+                self.cp_predicted_da = np.insert(self.cp_predicted_da, 0, 0)
+            
+            else:
+            
+                self.cp_predicted_da  = self.cp_da_nn[self.cosmo_model].predictions_np(params_dict)[0]
+            
+
+            self.chi_interp = scipy.interpolate.interp1d(
+                                                        self.cp_z_interp,
+                                                        self.cp_predicted_da*(1.+self.cp_z_interp),
+                                                        kind='linear',
+                                                        axis=-1,
+                                                        copy=True,
+                                                        bounds_error=None,
+                                                        fill_value=np.nan,
+                                                        assume_sorted=False)
 
     def get_cmb_cls(self,ell_factor=True,Tcmb_uk = Tcmb_uk):
 
         cls = {}
-        cls['ell'] = np.arange(self.cszfast_ldim)
-        cls['tt'] = np.zeros(self.cszfast_ldim)
-        cls['te'] = np.zeros(self.cszfast_ldim)
-        cls['ee'] = np.zeros(self.cszfast_ldim)
-        cls['pp'] = np.zeros(self.cszfast_ldim)
-        cls['bb'] = np.zeros(self.cszfast_ldim)
+        cls['ell'] = self.arange(self.cszfast_ldim)
+        cls['tt'] = self.zeros(self.cszfast_ldim)
+        cls['te'] = self.zeros(self.cszfast_ldim)
+        cls['ee'] = self.zeros(self.cszfast_ldim)
+        cls['pp'] = self.zeros(self.cszfast_ldim)
+        cls['bb'] = self.zeros(self.cszfast_ldim)
         cls['tt'][2:self.cp_lmax+1] = (Tcmb_uk)**2.*self.cp_predicted_tt_spectrum.copy()
         cls['te'][2:self.cp_lmax+1] = (Tcmb_uk)**2.*self.cp_predicted_te_spectrum.copy()
         cls['ee'][2:self.cp_lmax+1] = (Tcmb_uk)**2.*self.cp_predicted_ee_spectrum.copy()
@@ -831,12 +848,12 @@ class Class_szfast(object):
 
 
         if ell_factor==False:
-            fac_l = np.zeros(self.cszfast_ldim)
-            fac_l[2:self.cp_lmax+1] = 1./(cls['ell'][2:self.cp_lmax+1]*(cls['ell'][2:self.cp_lmax+1]+1.)/2./np.pi)
+            fac_l = self.zeros(self.cszfast_ldim)
+            fac_l[2:self.cp_lmax+1] = 1./(cls['ell'][2:self.cp_lmax+1]*(cls['ell'][2:self.cp_lmax+1]+1.)/2./self.pi)
             cls['tt'][2:self.cp_lmax+1] *= fac_l[2:self.cp_lmax+1]
             cls['te'][2:self.cp_lmax+1] *= fac_l[2:self.cp_lmax+1]
             cls['ee'][2:self.cp_lmax+1] *= fac_l[2:self.cp_lmax+1]
-            # cls['bb'] = np.zeros(self.cszfast_ldim)
+            # cls['bb'] = self.zeros(self.cszfast_ldim)
         return cls
 
 
@@ -882,7 +899,10 @@ class Class_szfast(object):
             return np.array(self.hz_interp(z)*H_units_conv_factor[units])
 
     def get_chi(self, z):
-        return np.array(self.chi_interp(z))
+        if self.jax_mode:
+            return jnp.array(self.chi_interp(z))
+        else:
+            return np.array(self.chi_interp(z))
 
     def get_gas_pressure_profile_x(self,z,m,x):
         return 0#np.vectorize(self.csz_base.get_pressure_P_over_P_delta_at_x_M_z_b12_200c)(x,m,z)
@@ -896,7 +916,7 @@ class Class_szfast(object):
 
 
     def tabulate_gas_pressure_profile_k(self):
-        z_asked,m_asked,x_asked = 0.2,3e14,np.geomspace(1e-3,1e2,500)
+        z_asked,m_asked,x_asked = 0.2,3e14,self.geomspace(1e-3,1e2,500)
         start = time.time()
         px = self.get_gas_pressure_profile_x(z_asked,m_asked,x_asked)
         end = time.time()

classy_szfast/cosmopower_jax.py

@@ -1,5 +1,6 @@
 from .config import path_to_class_sz_data
 import numpy as np
+import jax.numpy as jnp
 from .restore_nn import Restore_NN
 from .restore_nn import Restore_PCAplusNN
 from .suppress_warnings import suppress_warnings
@@ -20,6 +21,74 @@ cp_h_nn_jax = {}
 cp_s8_nn_jax = {}
 
 
+class CosmoPowerJAX_custom(CPJ):
+    def __init__(self, *args, **kwargs):
+        super().__init__(*args, **kwargs)
+        self.ten_to_predictions = True
+        if 'ten_to_predictions' in kwargs.keys():
+            self.ten_to_predictions = kwargs['ten_to_predictions']
+
+    def _predict(self, weights, hyper_params, param_train_mean, param_train_std,
+                 feature_train_mean, feature_train_std, input_vec):
+        """ Forward pass through pre-trained network.
+        In its current form, it does not make use of high-level frameworks like
+        FLAX et similia; rather, it simply loops over the network layers.
+        In future work this can be improved, especially if speed is a problem.
+        
+        Parameters
+        ----------
+        weights : array
+            The stored weights of the neural network.
+        hyper_params : array
+            The stored hyperparameters of the activation function for each layer.
+        param_train_mean : array
+            The stored mean of the training cosmological parameters.
+        param_train_std : array
+            The stored standard deviation of the training cosmological parameters.
+        feature_train_mean : array
+            The stored mean of the training features.
+        feature_train_std : array
+            The stored  standard deviation of the training features.
+        input_vec : array of shape (n_samples, n_parameters) or (n_parameters)
+            The cosmological parameters given as input to the network.
+            
+        Returns
+        -------
+        predictions : array
+            The prediction of the trained neural network.
+        """        
+        act = []
+        # Standardise
+        layer_out = [(input_vec - param_train_mean)/param_train_std]
+
+        # Loop over layers
+        for i in range(len(weights[:-1])):
+            w, b = weights[i]
+            alpha, beta = hyper_params[i]
+            act.append(jnp.dot(layer_out[-1], w.T) + b)
+            layer_out.append(self._activation(act[-1], alpha, beta))
+
+        # Final layer prediction (no activations)
+        w, b = weights[-1]
+        if self.probe == 'custom_log' or self.probe == 'custom_pca':
+            # in original CP models, we assumed a full final bias vector...
+            preds = jnp.dot(layer_out[-1], w.T) + b
+        else:   
+            # ... unlike in cpjax, where we used only a single bias vector
+            preds = jnp.dot(layer_out[-1], w.T) + b[-1]
+
+        # Undo the standardisation
+        preds = preds * feature_train_std + feature_train_mean
+        if self.log == True:
+            if self.ten_to_predictions:
+                preds = 10**preds
+        else:
+            preds = (preds@self.pca_matrix)*self.training_std + self.training_mean
+            if self.probe == 'cmb_pp':
+                preds = 10**preds
+        predictions = preds.squeeze()
+        return predictions
+
 for mp in cosmo_model_list:
     folder, version = split_emulator_string(mp)
     # print(folder, version)
@@ -36,18 +105,24 @@ for mp in cosmo_model_list:
     
     cp_pknl_nn_jax[mp] = Restore_NN(restore_filename=path_to_emulators + 'PK/' + emulator_dict[mp]['PKNL'])
     
-    cp_pkl_nn_jax[mp] = Restore_NN(restore_filename=path_to_emulators + 'PK/' + emulator_dict[mp]['PKL'])
-    
+    cp_pkl_nn_jax[mp] = CosmoPowerJAX_custom(probe='custom_log',filepath=path_to_emulators +'PK/' + emulator_dict[mp]['PKL'] + '.npz')
+    cp_pkl_nn_jax[mp].ten_to_predictions = False
+
     cp_der_nn_jax[mp] = Restore_NN(restore_filename=path_to_emulators + 'derived-parameters/' + emulator_dict[mp]['DER'])
     
-    cp_da_nn_jax[mp] = Restore_NN(restore_filename=path_to_emulators + 'growth-and-distances/' + emulator_dict[mp]['DAZ'])
+    # cp_da_nn_jax[mp] = Restore_NN(restore_filename=path_to_emulators + 'growth-and-distances/' + emulator_dict[mp]['DAZ'])
+    
     
+    cp_da_nn_jax[mp] = CosmoPowerJAX_custom(probe='custom_log',filepath=path_to_emulators + 'growth-and-distances/' + emulator_dict[mp]['DAZ'] + '.npz')
+    if mp != 'ede-v2':
+        cp_da_nn_jax[mp].ten_to_predictions = False
+    # print(cp_da_nn_jax[mp].parameters)
     # print(path_to_emulators + 'growth-and-distances/' + emulator_dict[mp]['HZ'])
-    emulator_custom = CPJ(probe='custom_log',filepath=path_to_emulators + 'growth-and-distances/' + emulator_dict[mp]['HZ'] + '.npz')
+    # emulator_custom = CPJ(probe='custom_log',filepath=path_to_emulators + 'growth-and-distances/' + emulator_dict[mp]['HZ'] + '.npz')
     # print(emulator_custom.parameters)
     # exit()
 
-    cp_h_nn_jax[mp] = CPJ(probe='custom_log',filepath=path_to_emulators + 'growth-and-distances/' + emulator_dict[mp]['HZ'] + '.npz')
+    cp_h_nn_jax[mp] = CosmoPowerJAX_custom(probe='custom_log',filepath=path_to_emulators + 'growth-and-distances/' + emulator_dict[mp]['HZ'] + '.npz')
 
     cp_s8_nn_jax[mp] = Restore_NN(restore_filename=path_to_emulators + 'growth-and-distances/' + emulator_dict[mp]['S8Z'])
     

classy_szfast/emulators_meta_data.py

@@ -33,7 +33,7 @@ cosmopower_derived_params_names = ['100*theta_s',
                                    'ra_star',
                                    'rs_drag']
 
-cp_l_max_scalars = 11000 # max multipole of train ing data
+cp_l_max_scalars = 11000 # max multipole of training data
 
 cosmo_model_list = [
     'lcdm',

{classy_szfast-0.0.25.post2.dist-info → classy_szfast-0.0.25.post4.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: classy_szfast
-Version: 0.0.25.post2
+Version: 0.0.25.post4
 Summary: The accelerator of the class_sz code from https://github.com/CLASS-SZ
 Maintainer-email: Boris Bolliet <bb667@cam.ac.uk>
 License: MIT

{classy_szfast-0.0.25.post2.dist-info → classy_szfast-0.0.25.post4.dist-info}/RECORD

@@ -1,11 +1,11 @@
 classy_szfast/__init__.py,sha256=E2thrL0Z9oXFfdzwcsu-xbOytudLFTlRlPqVFGlPPPg,279
 classy_szfast/classy_sz.py,sha256=QmbwrSXInQLMvCDqsr7KPmtaU0KOiOt1Rb-cTKuulZw,22240
-classy_szfast/classy_szfast.py,sha256=2eRVEjpTY9xkbHtIUZjZjKQJEcT_0-5ku0n63VYlXLo,36639
+classy_szfast/classy_szfast.py,sha256=H3DPD4_ZvVJ_FCVbEzc4rW8DYQHPZCCaGQZtQ1f-7oQ,37646
 classy_szfast/config.py,sha256=cd7Z62-qnX_4FJWfUNqcyJVh-AdBiXrF8DcQGpyAUZM,274
 classy_szfast/cosmopower.py,sha256=ooYK2BDOZSo3XtGHfPtjXHxr5UW-yVngLPkb5gpvTx8,2351
-classy_szfast/cosmopower_jax.py,sha256=C7NzfMFs9sL8rKuDdXdmwxk0UzHqNJnVjZENak-EPQA,2151
+classy_szfast/cosmopower_jax.py,sha256=NvwY1a9YoizOV0zKp0SEtiHQkhWRO-seGFzO3FIzLl0,5436
 classy_szfast/cosmosis_classy_szfast_interface.py,sha256=zAnxvFtn73a5yS7jgs59zpWFEYKCIQyraYPs5hQ4Le8,11483
-classy_szfast/emulators_meta_data.py,sha256=-lHneGhSJ2481S48viz_bNeCyAGu1Ogee0jFEB8B618,9724
+classy_szfast/emulators_meta_data.py,sha256=faA0iQqzfC5lOj1hu7wSEZUFfhBGLJqX9-xovMfTbr0,9723
 classy_szfast/pks_and_sigmas.py,sha256=drtuujE1HhlrYY1hY92DyY5lXlYS1uE15MSuVI4uo6k,6625
 classy_szfast/restore_nn.py,sha256=DqA9thhTRiGBDVb9zjhqcbF2W4V0AU0vrjJFhnLboU4,21075
 classy_szfast/suppress_warnings.py,sha256=6wIBml2Sj9DyRGZlZWhuA9hqvpxqrNyYjuz6BPK_a6E,202
@@ -14,7 +14,7 @@ classy_szfast/custom_bias/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJW
 classy_szfast/custom_bias/custom_bias.py,sha256=aR2t5RTIwv7P0m2bsEU0Eq6BTkj4pG10AebH6QpG4qM,486
 classy_szfast/custom_profiles/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 classy_szfast/custom_profiles/custom_profiles.py,sha256=4LZwb2XoqwCyWNmW2s24Z7AJdmgVdaRG7yYaBYe-d9Q,1188
-classy_szfast-0.0.25.post2.dist-info/METADATA,sha256=SHRC4v8N4uo_BSIHn9D8_KrReCq3coWsZCQIpLqaKmQ,548
-classy_szfast-0.0.25.post2.dist-info/WHEEL,sha256=P9jw-gEje8ByB7_hXoICnHtVCrEwMQh-630tKvQWehc,91
-classy_szfast-0.0.25.post2.dist-info/top_level.txt,sha256=hRgqpilUck4lx2KkaWI2y9aCDKqF6pFfGHfNaoPFxv0,14
-classy_szfast-0.0.25.post2.dist-info/RECORD,,
+classy_szfast-0.0.25.post4.dist-info/METADATA,sha256=oNj9IJ8zELtcPCEOOi4wQzx9SfxVaMmdIl6QjyAFdkY,548
+classy_szfast-0.0.25.post4.dist-info/WHEEL,sha256=P9jw-gEje8ByB7_hXoICnHtVCrEwMQh-630tKvQWehc,91
+classy_szfast-0.0.25.post4.dist-info/top_level.txt,sha256=hRgqpilUck4lx2KkaWI2y9aCDKqF6pFfGHfNaoPFxv0,14
+classy_szfast-0.0.25.post4.dist-info/RECORD,,