tensors used as indices must be long, byte or bool tensors

[yzj19870824]

error occurs when I train model using UMLS dataset

Epoch 2: average training loss = -0.22322563640773296 entropy = 4.2052984714508055
=> saving checkpoint to '/content/MultiHopKG-master/model/umls-point.rs.conve-xavier-n/a-200-200-3-0.001-0.1-0.0-0.95-400-0.05/checkpoint-2.tar'
0% 0/11 [00:00<?, ?it/s]
Traceback (most recent call last):
File "/usr/lib/python3.7/runpy.py", line 193, in _run_module_as_main
"main", mod_spec)
File "/usr/lib/python3.7/runpy.py", line 85, in _run_code
exec(code, run_globals)
File "/content/MultiHopKG-master/src/experiments.py", line 765, in
run_experiment(args)
File "/content/MultiHopKG-master/src/experiments.py", line 746, in run_experiment
train(lf)
File "/content/MultiHopKG-master/src/experiments.py", line 235, in train
lf.run_train(train_data, dev_data)
File "/content/MultiHopKG-master/src/learn_framework.py", line 146, in run_train
dev_scores = self.forward(dev_data, verbose=False)
File "/content/MultiHopKG-master/src/learn_framework.py", line 202, in forward
pred_score = self.predict(mini_batch, verbose=verbose)
File "/content/MultiHopKG-master/src/rl/graph_search/pg.py", line 226, in predict
pn, e1, r, e2, kg, self.num_rollout_steps, self.beam_size)
File "/content/MultiHopKG-master/src/rl/graph_search/beam_search.py", line 168, in beam_search
pn.update_path(action, kg, offset=action_offset)
File "/content/MultiHopKG-master/src/rl/graph_search/pn.py", line 187, in update_path
offset_path_history(self.path, offset)
File "/content/MultiHopKG-master/src/rl/graph_search/pn.py", line 176, in offset_path_history
new_tuple = tuple([_x[:, offset, :] for _x in x])
File "/content/MultiHopKG-master/src/rl/graph_search/pn.py", line 176, in
new_tuple = tuple([_x[:, offset, :] for _x in x])
IndexError: tensors used as indices must be long, byte or bool tensors

Does anyone meet that?

[yzj19870824]

The type of elements in offset is float indeed.

[seanPudding]

search for action_beam_offset = action_ind / action_sapce_size and top_unique_beam_offset = top_unique_idx / action_sapce_size
and then modify them to action_beam_offset = action_ind // action_sapce_size and top_unique_beam_offset = top_unique_idx // action_sapce_size
This change can works and I see it in DacKGR, but I am not sure if it's right.

[wenjian-sc]

搜索action_beam_offset = action_ind / action_sapce_size和top_unique_beam_offset = top_unique_idx / action_sapce_size然后将它们修改为action_beam_offset = action_ind // action_sapce_size和top_unique_beam_offset = top_unique_idx // action_sapce_size 此更改可以工作，我在DacKGR中看到它，但我不确定它是否正确。

Which file is this line of code in? Do you mind telling me? I can't find this line of code.
action_beam_offset = action_ind / action_sapce_size和top_unique_beam_offset = top_unique_idx / action_sapce_size

[seanPudding]

action_beam_offset = action_ind / action_sapce_size (src/rl/graph_search/beam_search.py line:62)
top_unique_beam_offset = top_unique_idx / action_sapce_size (src/rl/graph_search/beam_search.py line:105)

[wenjian-sc]

action_beam_offset = action_ind / action_sapce_size （src/rl/graph_search/beam_search.py 行：62） top_unique_beam_offset = top_unique_idx / action_sapce_size （src/rl/graph_search/beam_search.py 行：105）

Thank you for your help. I found it. Thank you