def main(): USE_GPU = True if USE_GPU and torch.cuda.is_available(): torch.cuda.empty_cache() device = torch.device('cuda') else: device = torch.device('cpu') print('using device:', device) dtype = torch.float32 ''' filename = 'curr_model_soumya_mat_it_es' f = open(filename, 'rb') model = pickle.load(f) f.close() ''' n_epochs = 5 n_refinement = 5 batch_size = 32 model = UMWE(dtype, device, batch_size, n_epochs, n_refinement) model.build_model() model.discrim_fit() filename = 'curr_model_soumya_mat_it_es' f = open(filename, 'wb') pickle.dump(model, f) f.close() eval_ = Evaluator(model) print(eval_.clws('es', 'en')) print(eval_.clws('en', 'es')) print(eval_.clws('it', 'en')) eval_.word_translation('es', 'en') eval_.word_translation('en', 'es') eval_.word_translation('it', 'en') model.mpsr_refine() print(eval_.clws('es', 'en')) print(eval_.clws('en', 'es')) print(eval_.clws('it', 'en')) eval_.word_translation('es', 'en') eval_.word_translation('en', 'es') eval_.word_translation('it', 'en') filename = 'curr_model_soumya_mpsr_it_es' f = open(filename, 'wb') pickle.dump(model, f) f.close() for lang in model.src_langs.values(): model.export_embeddings(lang, model.embs, "txt", "20th")
def main(): USE_GPU = True if USE_GPU and torch.cuda.is_available(): torch.cuda.empty_cache() device = torch.device('cuda') else: device = torch.device('cpu') print('using device:', device) dtype = torch.float32 # ============================================================================= # filename = 'curr_model' # f = open(filename, 'rb') # model = pickle.load(f) # f.close() # # ============================================================================= model = UMWE(dtype, device, 32, 2) model.build_model() # model.discrim_fit() # filename = 'curr_model' # f = open(filename, 'wb') # pickle.dump(model, f) # f.close() # ============================================================================= model.mpsr_refine() # ============================================================================= # ============================================================================= # for lang in model.src_langs.values(): # model.export_embeddings(lang, model.embs, "txt") # ============================================================================= model.export_embeddings('es', model.embs, "txt") eval_ = Evaluator(model) print(eval_.clws('es', 'en')) eval_.word_translation('es', 'en')