def main(): reader1 = our_reader.Reader(cfg['vocab_size'], cfg['buckets']) reader2 = cornell_reader.Reader(cfg) reader1.build_dict(cfg['dictionary_name'], cfg['reversed_dictionary_name'], cfg['path']['train']) reader1.read_data(cfg['path']['train']) encoder_inputs_10 = [i for i in reader.dataset_enc if len(i) == 10] decoder_inputs_10 = [i for i in reader.dataset_dec if len(i) == 10] encoder_inputs__toks_10 = [i for i in reader.dataset_enc_tok if len(i) == 10] decoder_inputs__toks_10 = [i for i in reader.dataset_dec_tok if len(i) == 10] print("done") model.create_placeholders()
identifier = 'mnistfull' settings = utils.load_settings_from_file(identifier) samples, pdf, labels = data_utils.get_samples_and_labels(settings) locals().update(settings) # json.dump(settings, open('./experiments/settings/' + identifier + '.txt', 'w'), indent=0) data_path = './experiments/data/' + identifier + '.data.npy' np.save(data_path, {'samples': samples, 'pdf': pdf, 'labels': labels}) print('Saved training data to', data_path) # --- build model --- # Z, X, CG, CD, CS = model.create_placeholders(batch_size, seq_length, latent_dim, num_signals, cond_dim) discriminator_vars = [ 'hidden_units_d', 'seq_length', 'cond_dim', 'batch_size', 'batch_mean' ] discriminator_settings = dict((k, settings[k]) for k in discriminator_vars) generator_vars = [ 'hidden_units_g', 'seq_length', 'batch_size', 'num_generated_features', 'cond_dim', 'learn_scale' ] generator_settings = dict((k, settings[k]) for k in generator_vars) CGAN = (cond_dim > 0) print(CGAN) D_loss, G_loss, accuracy = model.GAN_loss(Z,
num_variables = samples.shape[2] print('num_variables:', num_variables) # --- save settings, data --- # print('Ready to run with settings:') for (k, v) in settings.items(): print(v, '\t', k) # add the settings to local environment # WARNING: at this point a lot of variables appear locals().update(settings) json.dump(settings, open('./experiments/settings/' + identifier + '.txt', 'w'), indent=0) # --- build model --- # # preparation: data placeholders and model parameters Z, X, T = model.create_placeholders(batch_size, seq_length, latent_dim, num_variables) discriminator_vars = [ 'hidden_units_d', 'seq_length', 'batch_size', 'batch_mean' ] discriminator_settings = dict((k, settings[k]) for k in discriminator_vars) generator_vars = ['hidden_units_g', 'seq_length', 'batch_size', 'learn_scale'] generator_settings = dict((k, settings[k]) for k in generator_vars) generator_settings['num_signals'] = num_variables # model: GAN losses D_loss, G_loss = model.GAN_loss(Z, X, generator_settings, discriminator_settings) D_solver, G_solver, priv_accountant = model.GAN_solvers( D_loss, G_loss, learning_rate,
# --- training sample --- # # --- save settings, data --- # print('Ready to run with settings:') for (k, v) in settings.items(): print(v, '\t', k) # add the settings to local environment # WARNING: at this point a lot of variables appear locals().update(settings) json.dump(settings, open('./experiments/settings/' + identifier + '.txt', 'w'), indent=0) # --- build model --- # Z, X, T = model.create_placeholders(batch_size, seq_length, latent_dim, num_signals) discriminator_vars = [ 'hidden_units_d', 'seq_length', 'batch_size', 'batch_mean' ] discriminator_settings = dict((k, settings[k]) for k in discriminator_vars) generator_vars = [ 'hidden_units_g', 'seq_length', 'batch_size', 'num_generated_features', 'learn_scale' ] generator_settings = dict((k, settings[k]) for k in generator_vars) D_loss, G_loss = model.GAN_loss(Z, X, generator_settings, discriminator_settings) D_solver, G_solver, priv_accountant = model.GAN_solvers(
# WARNING: at this point a lot of variables appear locals().update(settings) json.dump(settings, open('./experiments/settings/' + identifier + '.txt', 'w'), indent=0) epoch = 150 #if not data == 'load': # data_path = './experiments/data/' + identifier + '.data.npy' # np.save(data_path, {'samples': samples, 'pdf': pdf, 'labels': labels}) # print('Saved training data to', data_path) # --- build model --- # Z, X, CG, CD, CS = model.create_placeholders(batch_size, seq_length, latent_dim, num_generated_features, cond_dim) discriminator_vars = [ 'hidden_units_d', 'seq_length', 'cond_dim', 'batch_size', 'batch_mean', 'latent_dim' ] discriminator_settings = dict((k, settings[k]) for k in discriminator_vars) generator_vars = [ 'hidden_units_g', 'seq_length', 'batch_size', 'latent_dim', 'num_generated_features', 'cond_dim', 'learn_scale' ] generator_settings = dict((k, settings[k]) for k in generator_vars) CGAN = (cond_dim > 0) if CGAN: assert not predict_labels
cond_path = './experiments/cond_data/' + identifier + '.data.npy' np.save(cond_path, {'cond_samples': cond_samples_train}) print('Saved cond images to ', cond_path) # --- build model --- # if generate_test: eval.sine_plot(identifier, 250) exit() if cond_dim > 0: num_signals = 1 if info: latent_C, Z, X, CG, CD, CS, cond_sine = model.create_placeholders( batch_size, seq_length, latent_dim, num_signals, cond_dim, info, latent_C_dim) else: Z, X, CG, CD, CS, cond_sine = model.create_placeholders( batch_size, seq_length, latent_dim, num_signals, cond_dim, info) latent_C = None discriminator_vars = [ 'hidden_units_d', 'seq_length', 'cond_dim', 'batch_size', 'batch_mean' ] discriminator_settings = dict((k, settings[k]) for k in discriminator_vars) generator_vars = [ 'hidden_units_g', 'seq_length', 'batch_size', 'num_generated_features', 'cond_dim', 'learn_scale' ] generator_settings = dict((k, settings[k]) for k in generator_vars)