alternate loss functions

[topepo]

For each model, have a list of acceptable loss functions and an argument to switch.

> ls(pattern = "loss$", envir = asNamespace("torch"))
 [1] "nn_adaptive_log_softmax_with_loss"     "nn_bce_loss"                          
 [3] "nn_bce_with_logits_loss"               "nn_cosine_embedding_loss"             
 [5] "nn_cross_entropy_loss"                 "nn_ctc_loss"                          
 [7] "nn_hinge_embedding_loss"               "nn_kl_div_loss"                       
 [9] "nn_l1_loss"                            "nn_loss"                              
[11] "nn_margin_ranking_loss"                "nn_mse_loss"                          
[13] "nn_multi_margin_loss"                  "nn_multilabel_margin_loss"            
[15] "nn_multilabel_soft_margin_loss"        "nn_nll_loss"                          
[17] "nn_poisson_nll_loss"                   "nn_smooth_l1_loss"                    
[19] "nn_soft_margin_loss"                   "nn_triplet_margin_loss"               
[21] "nn_triplet_margin_with_distance_loss"  "nn_weighted_loss"                     
[23] "nnf_cosine_embedding_loss"             "nnf_ctc_loss"                         
[25] "nnf_hinge_embedding_loss"              "nnf_l1_loss"                          
[27] "nnf_margin_ranking_loss"               "nnf_mse_loss"                         
[29] "nnf_multi_margin_loss"                 "nnf_multilabel_margin_loss"           
[31] "nnf_multilabel_soft_margin_loss"       "nnf_nll_loss"                         
[33] "nnf_poisson_nll_loss"                  "nnf_smooth_l1_loss"                   
[35] "nnf_soft_margin_loss"                  "nnf_triplet_margin_loss"              
[37] "nnf_triplet_margin_with_distance_loss" "torch__ctc_loss"                      
[39] "torch__cudnn_ctc_loss"                 "torch__use_cudnn_ctc_loss"            
[41] "torch_cosine_embedding_loss"           "torch_cross_entropy_loss"             
[43] "torch_ctc_loss"                        "torch_hinge_embedding_loss"           
[45] "torch_huber_loss"                      "torch_l1_loss"                        
[47] "torch_margin_ranking_loss"             "torch_mse_loss"                       
[49] "torch_multi_margin_loss"               "torch_multilabel_margin_loss"         
[51] "torch_nll_loss"                        "torch_poisson_nll_loss"               
[53] "torch_smooth_l1_loss"                  "torch_soft_margin_loss"               
[55] "torch_triplet_margin_loss"