Colab demo not working

[Jurys22]

Hello,
I am trying to run your demo (with premium GPU or standard) but from the third block of code, something doesn't work.

`ValidationError Traceback (most recent call last)
in
18 text = "Did not enjoy the new Windows 8 and touchscreen functions ."
19 model = SpanModel(save_dir="pretrained_14lap", random_seed=0)
---> 20 sent = predict_sentence(text, model)
21
22 for t in sent.triples:

3 frames
in predict_sentence(text, model)
12 data = Data(root=Path(), data_split=SplitEnum.test, sentences=[sent])
13 data.save_to_path(path_in)
---> 14 model.predict(path_in, path_out)
15 data = Data.load_from_full_path(path_out)
16 return data.sentences[0]

/content/aste/wrapper.py in predict(self, path_in, path_out)
89
90 with open(path_temp_out) as f:
---> 91 preds = [SpanModelPrediction(**json.loads(line.strip())) for line in f]
92 data = Data(
93 root=Path(),

/content/aste/wrapper.py in (.0)
89
90 with open(path_temp_out) as f:
---> 91 preds = [SpanModelPrediction(**json.loads(line.strip())) for line in f]
92 data = Data(
93 root=Path(),

/usr/local/lib/python3.7/dist-packages/pydantic/main.cpython-37m-x86_64-linux-gnu.so in pydantic.main.BaseModel.init()

ValidationError: 1 validation error for SpanModelPrediction
predicted_relations
field required (type=value_error.missing)`

[chiayewken]

Hi, thanks for pointing this out, I have updated the version and it should be fixed in the colab now
https://colab.research.google.com/drive/1F9zW_nVkwfwIVXTOA_juFDrlPz5TLjpK?usp=sharing#scrollTo=izKXA4b6-oIv

[Jurys22]

Hi again!
I tried to launch it this morning, making a copy of the colab on my colab (using GPU) but it still gives me the error.

[Jurys22]

For info, if I run the colab directly from your link, without copying it on my drive it works. Any hints on why is this happening?

[chiayewken]

Hi, it may be a difference in environment, is your notebook using GPU or CPU?

[Jurys22]

I think I'm using a GPU (different one than yours)

[Jurys22]

When I'm on a Tesla T4 seems working. I don't know if there is a way to choose the GPU selectively to test with another type of GPU though. Any other reasons why this could happen?

[chiayewken]

I'm not sure about the reason for errors with specific GPUs, but I have updated the pretrained weights and demo notebooks, please git pull to latest commit and try again, thanks for raising the issue!

[Jurys22]

I think I found the issue - that it is still there.

_A100-SXM4-40GB with CUDA capability sm_80 is not compatible with the current PyTorch installation.
The current PyTorch install supports CUDA capabilities sm_37 sm_50 sm_60 sm_70 sm_75.
If you want to use the A100-SXM4-40GB GPU with PyTorch, please check the instructions at https://pytorch.org/get-started/locally/

warnings.warn(incompatible_device_warn.format(device_name, capability, " ".join(arch_list), device_name))
reading instances: 1it [00:00, 447.49it/s]
/content/span_model/predictors/span_model.py:69: UserWarning: Encountered a RunTimeError on document 0. Skipping this example. Error message:
CUDA error: no kernel image is available for execution on the device.
warnings.warn(msg)_