rnn: hx is not contiguous[BUG]

[korosig]

Hello,

Nice work, I like your package.

After a few days of waiting, I finally managed to install all the necessary book transfers on my computer.
When my Torch was running on CPU only, the templates ran fine, but now I wanted to test the program on GPU and got the following Error:

rnn: hx is not contiguous

Template:
https://github.com/h3ik0th/TFT_darts/blob/main/TFT_2g5.ipynb

System :
absl-py
anyio==3.4.0
argcomplete
argon2-cffi
argon2-cffi-bindings
arviz
astor
async-generator==1.10
attrs
Babel==2.9.1
backcall
backports.functools-lru-cache
bleach
bokeh
cached-property
cachetools==4.2.4
certifi==2021.10.8
cffi
cftime
charset-normalizer==2.0.10
click
cloudpickle
cmdstanpy==0.9.68
colorama
convertdate
cycler
Cython
cytoolz==0.11.2
darts==0.15.0
dask
dataclasses
debugpy
decorator
defusedxml
distributed
docopt==0.6.2
entrypoints
ephem
filterpy==1.4.5
flit_core
fonttools
fsspec
gast
google-auth==2.3.3
google-auth-oauthlib==0.4.6
google-pasta==0.2.0
grpcio
h5py
HeapDict==1.0.1
hijri-converter
holidays
idna==3.3
importlib-metadata
importlib-resources
ipykernel
ipython
ipython-genutils==0.2.0
ipywidgets
jedi
Jinja2
joblib==1.1.0
json5==0.9.6
jsonschema
jupyter-client
jupyter-core
jupyter-server==1.13.1
jupyterlab==3.2.6
jupyterlab-pygments
jupyterlab-server==2.10.3
jupyterlab-widgets
Keras-Applications==1.0.8
Keras-Preprocessing
kiwisolver
korean-lunar-calendar
lightgbm==3.3.2
locket==0.2.0
LunarCalendar==0.0.9
Markdown
MarkupSafe
matplotlib==3.5.1
matplotlib-inline
mistune
msgpack
munkres==1.1.4
nbclassic==0.3.4
nbclient
nbconvert
nbformat
nest-asyncio
netCDF4
notebook
numpy==1.21.5
oauthlib==3.1.1
olefile
packaging
pandas
pandocfilters
parso
partd
patsy==0.5.2
pickleshare
Pillow==9.0.0
pipreqs==0.4.11
pmdarima==1.8.4
prometheus-client rk
prompt-toolkit
prophet
protobuf==3.19.2
psutil
pyarrow==6.0.1
pyasn1==0.4.8
pyasn1-modules==0.2.8
pycparser
Pygments
PyMeeus==0.5.11
pyparsing
PyQt5==5.12.3
PyQt5_sip==4.19.18
PyQtChart==5.12
PyQtWebEngine==5.12.1
pyrsistent
pystan==2.19.1.1
python-dateutil
pytz
pywin32==303
pywinpty
PyYAML
pyzmq
requests==2.27.1
requests-oauthlib==1.3.0
rsa==4.8
scikit-learn==1.0.2
scipy
Send2Trash
setuptools-git==1.2
six
sniffio==1.2.0
sortedcontainers
statsmodels==0.13.1
tblib
tensorboard==1.15.0
tensorboard-data-server==0.6.1
tensorboard-plugin-wit==1.8.1
tensorflow 1.14l
tensorflow-estimator==1.14.0
termcolor==1.1.0
terminado
testpath
threadpoolctl==3.0.0
toolz
torch==1.10.1+cu113
torchaudio==0.10.1+cu113
torchvision==0.11.2+cu113
tornado
tqdm
traitlets
typing_extensions==4.0.1
ujson==5.1.0
unicodedata2
urllib3==1.26.8
wcwidth
webencodings==0.5.1
websocket-client==1.2.3
Werkzeug==2.0.2
widgetsnbextension k
wrapt
xarray
yarg==0.1.9
zict==2.0.0
zipp

Error :

RuntimeError Traceback (most recent call last)
~\AppData\Local\Temp/ipykernel_29804/2078052997.py in
2 model.fit( ts_ttrain,
3 future_covariates=tcov,
----> 4 verbose=True)

~\anaconda3\envs\TFT-test\lib\site-packages\darts\utils\torch.py in decorator(self, *args, **kwargs)
63 with fork_rng():
64 manual_seed(self._random_instance.randint(0, high=MAX_TORCH_SEED_VALUE))
---> 65 return decorated(self, *args, **kwargs)
66 return decorator

~\anaconda3\envs\TFT-test\lib\site-packages\darts\models\forecasting\torch_forecasting_model.py in fit(self, series, past_covariates, future_covariates, val_series, val_past_covariates, val_future_covariates, verbose, epochs, max_samples_per_ts, num_loader_workers)
477 logger.info('Train dataset contains {} samples.'.format(len(train_dataset)))
478
--> 479 self.fit_from_dataset(train_dataset, val_dataset, verbose, epochs, num_loader_workers)
480
481 @Property

~\anaconda3\envs\TFT-test\lib\site-packages\darts\utils\torch.py in decorator(self, *args, **kwargs)
63 with fork_rng():
64 manual_seed(self._random_instance.randint(0, high=MAX_TORCH_SEED_VALUE))
---> 65 return decorated(self, *args, **kwargs)
66 return decorator

~\anaconda3\envs\TFT-test\lib\site-packages\darts\models\forecasting\torch_forecasting_model.py in fit_from_dataset(self, train_dataset, val_dataset, verbose, epochs, num_loader_workers)
591
592 # Train model
--> 593 self._train(train_loader, val_loader, tb_writer, verbose, train_num_epochs)
594
595 # Close tensorboard writer

~\anaconda3\envs\TFT-test\lib\site-packages\darts\models\forecasting\torch_forecasting_model.py in _train(self, train_loader, val_loader, tb_writer, verbose, epochs)
886 self.model.train()
887 train_batch = self._batch_to_device(train_batch)
--> 888 output = self._produce_train_output(train_batch[:-1])
889 target = train_batch[-1] # By convention target is always the last element returned by datasets
890 loss = self._compute_loss(output, target)

~\anaconda3\envs\TFT-test\lib\site-packages\darts\models\forecasting\tft_model.py in _produce_train_output(self, input_batch)
800
801 def _produce_train_output(self, input_batch: Tuple):
--> 802 return self.model(input_batch)
803
804 def predict(self, n, *args, **kwargs):

~\anaconda3\envs\TFT-test\lib\site-packages\torch\nn\modules\module.py in _call_impl(self, *input, **kwargs)
1100 if not (self._backward_hooks or self._forward_hooks or self._forward_pre_hooks or _global_backward_hooks
1101 or _global_forward_hooks or _global_forward_pre_hooks):
-> 1102 return forward_call(*input, **kwargs)
1103 # Do not call functions when jit is used
1104 full_backward_hooks, non_full_backward_hooks = [], []

~\anaconda3\envs\TFT-test\lib\site-packages\darts\models\forecasting\tft_model.py in forward(self, x)
447
448 # run local lstm encoder
--> 449 encoder_out, (hidden, cell) = self.lstm_encoder(input=embeddings_varying_encoder, hx=(input_hidden, input_cell))
450
451 # run local lstm decoder

~\anaconda3\envs\TFT-test\lib\site-packages\torch\nn\modules\module.py in _call_impl(self, *input, **kwargs)
1100 if not (self._backward_hooks or self._forward_hooks or self._forward_pre_hooks or _global_backward_hooks
1101 or _global_forward_hooks or _global_forward_pre_hooks):
-> 1102 return forward_call(*input, **kwargs)
1103 # Do not call functions when jit is used
1104 full_backward_hooks, non_full_backward_hooks = [], []

~\anaconda3\envs\TFT-test\lib\site-packages\torch\nn\modules\rnn.py in forward(self, input, hx)
690 if batch_sizes is None:
691 result = _VF.lstm(input, hx, self._flat_weights, self.bias, self.num_layers,
--> 692 self.dropout, self.training, self.bidirectional, self.batch_first)
693 else:
694 result = _VF.lstm(input, batch_sizes, hx, self._flat_weights, self.bias,

RuntimeError: rnn: hx is not contiguous

[korosig]

actually I don't know if this is a good solution, but in TFT_model.py line 449 I overwrote the

run local lstm encoder
encoder_out, (hidden, cell) = self.lstm_encoder(input=embeddings_varying_encoder, hx=(input_hidden, input_cell))

to

run local lstm encoder
encoder_out, (hidden, cell) = self.lstm_encoder(input=embeddings_varying_encoder, hx=(input_hidden.contiguous(), input_cell.contiguous()))

[dennisbader]

Hi @korosig and thanks for writing.

We are working on it.
I close this as it is a duplicate of #722.