no drop for loader

domainlab/algos/trainers/a_trainer.py

@@ -36,6 +36,7 @@ def __init__(self, successor_node=None):
         self.aconf = None
         #
         self.loader_tr = None
+        self.loader_tr_no_drop = None
         self.loader_te = None
         self.num_batches = None
         self.flag_update_hyper_per_epoch = None
@@ -67,6 +68,7 @@ def init_business(self, model, task, observer, device, aconf, flag_accept=True):
         self.aconf = aconf
         #
         self.loader_tr = task.loader_tr
+        self.loader_tr_no_drop = task._loader_tr_no_drop
         self.loader_te = task.loader_te
 
         if flag_accept:

domainlab/algos/trainers/train_fbopt.py

@@ -71,9 +71,8 @@ def eval_p_loss(self, dict4mu, dict_theta):
         temp_model.hyper_update(epoch=None, fun_scheduler=HyperSetter(dict4mu))
         temp_model.set_params(dict_theta)
         epo_p_loss = 0  # penalized loss
-        # FIXME: will loader be corupted? if called at different places? if we do not make deep copy
         with torch.no_grad():
-            for _, (tensor_x, vec_y, vec_d, *_) in enumerate(self.loader_tr):
+            for _, (tensor_x, vec_y, vec_d, *_) in enumerate(self.loader_tr_no_drop):
                 tensor_x, vec_y, vec_d = \
                     tensor_x.to(self.device), vec_y.to(self.device), vec_d.to(self.device)
                 # sum will kill the dimension of the mini batch
@@ -87,14 +86,12 @@ def eval_r_loss(self):
         ERM loss on all available training data
         # TODO: normalize loss via batchsize
         """
-        # FIXME: move this to model instead of having it in trainer here
         temp_model = copy.deepcopy(self.model)
         # mock the model hyper-parameter to be from dict4mu
         epo_reg_loss = 0
         epo_task_loss = 0
-        # FIXME: will loader be corupted? if called at different places? if we do not make deep copy
         with torch.no_grad():
-            for _, (tensor_x, vec_y, vec_d, *_) in enumerate(self.loader_tr):
+            for _, (tensor_x, vec_y, vec_d, *_) in enumerate(self.loader_tr_no_drop):
                 tensor_x, vec_y, vec_d = \
                     tensor_x.to(self.device), vec_y.to(self.device), vec_d.to(self.device)
                 b_reg_loss = temp_model.cal_reg_loss(tensor_x, vec_y, vec_d).sum()

domainlab/tasks/a_task.py

@@ -16,6 +16,7 @@ class NodeTaskDG(AbstractChainNodeHandler):
     def __init__(self, succ=None):
         super().__init__(succ)
         self._loader_tr = None
+        self._loader_tr_no_drop = None
         self._loader_te = None
         self._loader_val = None
         self._list_domains = None

domainlab/tasks/b_task.py

@@ -43,6 +43,7 @@ def init_business(self, args, node_algo_builder=None):
             self.dict_dset_val.update({na_domain: ddset_val})
         ddset_mix = ConcatDataset(tuple(self.dict_dset_tr.values()))
         self._loader_tr = mk_loader(ddset_mix, args.bs)
+        self._loader_tr_no_drop = mk_loader(ddset_mix, args.bs, drop_last=False, shuffle=False)
 
         ddset_mix_val = ConcatDataset(tuple(self.dict_dset_val.values()))
         self._loader_val = mk_loader(ddset_mix_val, args.bs,