Mercurial > ift6266

comparison deep/stacked_dae/v_sylvain/nist_sda_retrieve.py @ 313:e301a2f32665

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Avoir exactement le meme jeu de donnees pour pre-train et finetune
author SylvainPL <sylvain.pannetier.lebeuf@umontreal.ca>
date Thu, 01 Apr 2010 14:25:55 -0400
parents 8b31280129a9
children 60e82846a10d
comparison
equal deleted inserted replaced
312:bd6085d77706 313:e301a2f32665
116 #Decide how the finetune is done 116 #Decide how the finetune is done
117 117
118 if finetune_choice == 0: 118 if finetune_choice == 0:
119 print('\n\n\tfinetune with NIST\n\n') 119 print('\n\n\tfinetune with NIST\n\n')
120 optimizer.reload_parameters(PATH+'params_pretrain.txt') 120 optimizer.reload_parameters(PATH+'params_pretrain.txt')
121 optimizer.finetune(datasets.nist_all(),datasets.nist_P07(min_file=nb_file),max_finetune_epoch_NIST,ind_test=1) 121 optimizer.finetune(datasets.nist_all(),datasets.nist_P07(),max_finetune_epoch_NIST,ind_test=1)
122 channel.save() 122 channel.save()
123 if finetune_choice == 1: 123 if finetune_choice == 1:
124 print('\n\n\tfinetune with P07\n\n') 124 print('\n\n\tfinetune with P07\n\n')
125 optimizer.reload_parameters(PATH+'params_pretrain.txt') 125 optimizer.reload_parameters(PATH+'params_pretrain.txt')
126 optimizer.finetune(datasets.nist_P07(min_file=nb_file),datasets.nist_all(),max_finetune_epoch_P07,ind_test=0) 126 optimizer.finetune(datasets.nist_P07(),datasets.nist_all(),max_finetune_epoch_P07,ind_test=0)
127 channel.save() 127 channel.save()
128 if finetune_choice == 2: 128 if finetune_choice == 2:
129 print('\n\n\tfinetune with NIST followed by P07\n\n') 129 print('\n\n\tfinetune with NIST followed by P07\n\n')
130 optimizer.reload_parameters(PATH+'params_pretrain.txt') 130 optimizer.reload_parameters(PATH+'params_pretrain.txt')
131 optimizer.finetune(datasets.nist_all(),datasets.nist_P07(min_file=nb_file),max_finetune_epoch_NIST,ind_test=21) 131 optimizer.finetune(datasets.nist_all(),datasets.nist_P07(),max_finetune_epoch_NIST,ind_test=21)
132 optimizer.finetune(datasets.nist_P07(min_file=nb_file),datasets.nist_all(),max_finetune_epoch_P07,ind_test=20) 132 optimizer.finetune(datasets.nist_P07(),datasets.nist_all(),max_finetune_epoch_P07,ind_test=20)
133 channel.save() 133 channel.save()
134 if finetune_choice == 3: 134 if finetune_choice == 3:
135 print('\n\n\tfinetune with NIST only on the logistic regression on top (but validation on P07).\n\ 135 print('\n\n\tfinetune with NIST only on the logistic regression on top (but validation on P07).\n\
136 All hidden units output are input of the logistic regression\n\n') 136 All hidden units output are input of the logistic regression\n\n')
137 optimizer.reload_parameters(PATH+'params_pretrain.txt') 137 optimizer.reload_parameters(PATH+'params_pretrain.txt')
138 optimizer.finetune(datasets.nist_all(),datasets.nist_P07(min_file=nb_file),max_finetune_epoch_NIST,ind_test=1,special=1) 138 optimizer.finetune(datasets.nist_all(),datasets.nist_P07(),max_finetune_epoch_NIST,ind_test=1,special=1)
139 139
140 140
141 if finetune_choice==-1: 141 if finetune_choice==-1:
142 print('\nSERIE OF 3 DIFFERENT FINETUNINGS') 142 print('\nSERIE OF 3 DIFFERENT FINETUNINGS')
143 print('\n\n\tfinetune with NIST\n\n') 143 print('\n\n\tfinetune with NIST\n\n')
144 sys.stdout.flush() 144 sys.stdout.flush()
145 optimizer.reload_parameters(PATH+'params_pretrain.txt') 145 optimizer.reload_parameters(PATH+'params_pretrain.txt')
146 optimizer.finetune(datasets.nist_all(),datasets.nist_P07(min_file=nb_file),max_finetune_epoch_NIST,ind_test=1) 146 optimizer.finetune(datasets.nist_all(),datasets.nist_P07(),max_finetune_epoch_NIST,ind_test=1)
147 channel.save() 147 channel.save()
148 print('\n\n\tfinetune with P07\n\n') 148 print('\n\n\tfinetune with P07\n\n')
149 sys.stdout.flush() 149 sys.stdout.flush()
150 optimizer.reload_parameters(PATH+'params_pretrain.txt') 150 optimizer.reload_parameters(PATH+'params_pretrain.txt')
151 optimizer.finetune(datasets.nist_P07(min_file=nb_file),datasets.nist_all(),max_finetune_epoch_P07,ind_test=0) 151 optimizer.finetune(datasets.nist_P07(),datasets.nist_all(),max_finetune_epoch_P07,ind_test=0)
152 channel.save() 152 channel.save()
153 print('\n\n\tfinetune with NIST (done earlier) followed by P07 (written here)\n\n') 153 print('\n\n\tfinetune with NIST (done earlier) followed by P07 (written here)\n\n')
154 sys.stdout.flush() 154 sys.stdout.flush()
155 optimizer.reload_parameters('params_finetune_NIST.txt') 155 optimizer.reload_parameters('params_finetune_NIST.txt')
156 optimizer.finetune(datasets.nist_P07(min_file=nb_file),datasets.nist_all(),max_finetune_epoch_P07,ind_test=20) 156 optimizer.finetune(datasets.nist_P07(),datasets.nist_all(),max_finetune_epoch_P07,ind_test=20)
157 channel.save() 157 channel.save()
158 print('\n\n\tfinetune with NIST only on the logistic regression on top.\n\ 158 print('\n\n\tfinetune with NIST only on the logistic regression on top.\n\
159 All hidden units output are input of the logistic regression\n\n') 159 All hidden units output are input of the logistic regression\n\n')
160 sys.stdout.flush() 160 sys.stdout.flush()
161 optimizer.reload_parameters(PATH+'params_pretrain.txt') 161 optimizer.reload_parameters(PATH+'params_pretrain.txt')
162 optimizer.finetune(datasets.nist_all(),datasets.nist_P07(min_file=nb_file),max_finetune_epoch_NIST,ind_test=1,special=1) 162 optimizer.finetune(datasets.nist_all(),datasets.nist_P07(),max_finetune_epoch_NIST,ind_test=1,special=1)
163 channel.save() 163 channel.save()
164 164
165 channel.save() 165 channel.save()
166 166
167 return channel.COMPLETE 167 return channel.COMPLETE