Issue while Training on CPU

[Ralfons-06]

Hi, I just tried to train the model on a cpu but i ran into some Problems.
While Training i always get the output message, that the loss at iteration x is 0 which seems kinda odd:

NLL Loss @ epoch 0001 iteration 00000001 = 0.0000
NLL Loss @ epoch 0063 iteration 00000250 = 0.0000

After going through the code of gran_runner i realized that the part of the code, where the loss is calculated is never called when there is no gpu available since batch_fwd is empty in that case:

GRAN/runner/gran_runner.py

Lines 230 to 259 in 43cb443

	avg_train_loss = .0
	for ff in range(self.dataset_conf.num_fwd_pass):
	batch_fwd = []
	if self.use_gpu:
	for dd, gpu_id in enumerate(self.gpus):
	data = {}
	data['adj'] = batch_data[dd][ff]['adj'].pin_memory().to(gpu_id, non_blocking=True)
	data['edges'] = batch_data[dd][ff]['edges'].pin_memory().to(gpu_id, non_blocking=True)
	data['node_idx_gnn'] = batch_data[dd][ff]['node_idx_gnn'].pin_memory().to(gpu_id, non_blocking=True)
	data['node_idx_feat'] = batch_data[dd][ff]['node_idx_feat'].pin_memory().to(gpu_id, non_blocking=True)
	data['label'] = batch_data[dd][ff]['label'].pin_memory().to(gpu_id, non_blocking=True)
	data['att_idx'] = batch_data[dd][ff]['att_idx'].pin_memory().to(gpu_id, non_blocking=True)
	data['subgraph_idx'] = batch_data[dd][ff]['subgraph_idx'].pin_memory().to(gpu_id, non_blocking=True)
	data['subgraph_idx_base'] = batch_data[dd][ff]['subgraph_idx_base'].pin_memory().to(gpu_id, non_blocking=True)
	batch_fwd.append((data,))
	if batch_fwd:
	train_loss = model(*batch_fwd).mean()
	avg_train_loss += train_loss
	# assign gradient
	train_loss.backward()
	# clip_grad_norm_(model.parameters(), 5.0e-0)
	optimizer.step()
	avg_train_loss /= float(self.dataset_conf.num_fwd_pass)
	# reduce
	train_loss = float(avg_train_loss.data.cpu().numpy())

Is this a bug or did i miss something?