uwnlp / open_type Goto Github PK

Hi,

I already produced a dataset in the same format as yours, but I noticed that the code is a little tricky.

How can I do experiments on my dataset?

Is there any way to predict entity types on new data?

I kept getting "TypeError: slice indices must be integers or None or have an index method" errors. I fixed it by casting floats to integers in models.py:

def sorted_rnn(self, sequences, sequence_lengths, rnn):
    sorted_inputs, sorted_sequence_lengths, restoration_indices = sort_batch_by_length(sequences, sequence_lengths)
    packed_sequence_input = pack_padded_sequence(sorted_inputs,
                                                 [int(_) for _ in sorted_sequence_lengths.data.tolist()],
                                                 batch_first=True)
    packed_sequence_output, _ = rnn(packed_sequence_input, None)
    unpacked_sequence_tensor, _ = pad_packed_sequence(packed_sequence_output, batch_first=True)
    return unpacked_sequence_tensor.index_select(0, restoration_indices)

Run command: python main.py onto -lstm_type single -goal onto -enhanced_mention

Error:

.../model_utils.py:167: UserWarning: Implicit dimension choice for softmax has been deprecated. Change the call to include dim=X as an argument.
weighted_keys = self.key_softmax(k).view(input_embed.size()[0], -1, 1)
Traceback (most recent call last):
File "main.py", line 276, in
_train(config)
File "main.py", line 133, in _train
total_loss += loss.data.cpu()[0]
IndexError: invalid index of a 0-dim tensor. Use tensor.item() to convert a 0-dim tensor to a Python number

-rw-rw-r-- 1 gpuws gpuws 5619697 12月 24 22:09 onto_95000.pt
-rw-rw-r-- 1 gpuws gpuws 5619697 12月 24 22:13 onto_96000.pt
-rw-rw-r-- 1 gpuws gpuws 5619697 12月 24 22:17 onto_97000.pt
-rw-rw-r-- 1 gpuws gpuws 5619697 12月 24 22:20 onto_98000.pt
-rw-rw-r-- 1 gpuws gpuws 5619697 12月 24 22:24 onto_99000.pt
(.venv) gpuws@gpuws32g:/media/gpuws/fcd84300-9270-4bbd-896a-5e04e79203b7/ub16_prj/open_type$ python3 main.py onto -lstm_type single -enhanced_mention -data_setup joint -add_crowd -multitask -mode test -reload_model_name onto_98000 -eval_data crowd/test.json -load
Loading word embeddings from /home/gpuws/data/glove/glove.840B.300d.txt...
Done loading word embeddings!
Answer num 10331
Found 1 shards at release/crowd/test.json
Multi-task learning
Traceback (most recent call last):
File "/media/gpuws/fcd84300-9270-4bbd-896a-5e04e79203b7/ub16_prj/open_type/.venv/lib/python3.5/site-packages/torch/nn/modules/module.py", line 514, in load_state_dict
own_state[name].copy_(param)
RuntimeError: invalid argument 2: sizes do not match at /pytorch/torch/lib/THC/THCTensorCopy.cu:31

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
File "main.py", line 277, in
_test(config)
File "main.py", line 230, in _test
load_model(args.reload_model_name, constant.EXP_ROOT, args.model_id, model)
File "main.py", line 203, in load_model
model.load_state_dict(checkpoint['state_dict'])
File "/media/gpuws/fcd84300-9270-4bbd-896a-5e04e79203b7/ub16_prj/open_type/.venv/lib/python3.5/site-packages/torch/nn/modules/module.py", line 519, in load_state_dict
.format(name, own_state[name].size(), param.size()))
RuntimeError: While copying the parameter named decoder.linear.weight, whose dimensions in the model are torch.Size([10331, 550]) and whose dimensions in the checkpoint are torch.Size([89, 550]).
(.venv) gpuws@gpuws32g:/media/gpuws/fcd84300-9270-4bbd-896a-5e04e79203b7/ub16_prj/open_type$
(.venv) gpuws@gpuws32g:/media/gpuws/fcd84300-9270-4bbd-896a-5e04e79203b7/ub16_prj/open_type$

The University of Washington works hard to foster an inclusive and welcoming culture for everyone in the campus community.

Unfortunately, looking over your GitHub repositories, I have identified language that can be considered offensive due to its racist origins. Specifically the branch names, but I imagine if I looked at the code itself, I would find more.

Can you let me know when you’ll be adjusting the settings to be a inclusive product/service and follow University of Washington language guidelines?

There is no mention of how vocabulary are obtained. Can you let me know?

Does the training stop at some point? The code says "while True". I hope it's not an infinite loop. Thanks.

Hi, the link provided in the Readme for the project website is not working.