From 5ecf211947135e77e183207b615b0a68d3c204ad Mon Sep 17 00:00:00 2001
From: William D'Arcy Kenworthy <wkenwor1@jh.edu>
Date: Wed, 3 Jul 2024 10:05:50 -0500
Subject: [PATCH] Accidentally removed outputdir from gradientdescent causing a
 crash Fixed an initialization bug which wrote intial x0 guess to x1 instead

---
 saltshaker/training/TrainSALT.py                  |  5 ++---
 saltshaker/training/base.py                       | 14 +++++++-------
 saltshaker/training/optimizers/gradientdescent.py |  1 +
 3 files changed, 10 insertions(+), 10 deletions(-)

diff --git a/saltshaker/training/TrainSALT.py b/saltshaker/training/TrainSALT.py
index 7e46d4fe..6d150e7d 100755
--- a/saltshaker/training/TrainSALT.py
+++ b/saltshaker/training/TrainSALT.py
@@ -336,7 +336,6 @@ def bestfit(p):
 
             from numpy.random import default_rng
             rng = default_rng(134912348)
-
             for sn in datadict.keys():
                 if self.options.snparlist:
                     # hacky matching, but SN names are a mess as usual
@@ -354,8 +353,8 @@ def bestfit(p):
                             else:
                                 guess[parlist==f'x{i}_{sn}'] = rng.standard_normal()
                         if snpar['x0'][iSN]<= 0:
-                            log.warning(f'Bad input value for {sn}: x0= {snpar["x0"][iSN]}')
-                            guess[parlist==f'x{i}_{sn}'] = 10**(-0.4*(cosmo.distmod(datadict[sn].zHelio).value-19.36-10.635))
+                            log.warning(f'Bad input value for {sn}: x0={ float(snpar["x0"][iSN])}')
+                            guess[parlist==f'x0_{sn}'] = 10**(-0.4*(cosmo.distmod(datadict[sn].zHelio).value-19.36-10.635))
 
                         guess[parlist == 'c0_%s'%sn] = snpar['c'][iSN]
                         guess[parlist == 'c1_%s'%sn] = np.random.exponential(0.2)
diff --git a/saltshaker/training/base.py b/saltshaker/training/base.py
index 99ec000b..7d73617b 100644
--- a/saltshaker/training/base.py
+++ b/saltshaker/training/base.py
@@ -322,14 +322,14 @@ def mkcuts(self,datadict):
             return outdict,cutdict
 
         def filter_select(self,survey,flt):
-                select = True
                 if flt in self.options.__dict__[f"{survey.split('(')[0]}_ignore_filters"].replace(' ','').split(','):
-                        select = False
+                    return  False
 
-                lambdaeff = self.kcordict[survey][flt]['lambdaeff']
-                if lambdaeff < self.options.filtercen_obs_waverange[0] or \
-                   lambdaeff > self.options.filtercen_obs_waverange[1] :
-                        select = False                
+                else:
+                    lambdaeff = self.kcordict[survey][flt]['lambdaeff']
+                    if lambdaeff < self.options.filtercen_obs_waverange[0] or \
+                                 lambdaeff > self.options.filtercen_obs_waverange[1] :
+                         return False
 
-                return select                
+                return True 
                 # end filter_select
diff --git a/saltshaker/training/optimizers/gradientdescent.py b/saltshaker/training/optimizers/gradientdescent.py
index 683a1697..bfdbecf5 100644
--- a/saltshaker/training/optimizers/gradientdescent.py
+++ b/saltshaker/training/optimizers/gradientdescent.py
@@ -47,6 +47,7 @@ def __init__(self,guess,saltresids,outputdir,options):
         self.saltobj=saltresids  
         for x in self.configoptionnames:
      	    self.__dict__[x]=getattr(options,x)
+        self.outputdir=options.outputdir
         assert(0<self.searchsize<1)
         assert(0<self.searchtolerance<1)
         assert(0<self.etaminus<1)