endrikacupaj / lasagne Goto Github PK

Hi~

This work is really great.

However, there is no mention in the paper or in the repository about what kind of GPU LASAGNE uses for training, so if you could provide me with some details about it, it would help me a lot.

cheers

Hello, thanks for your sharing and contributions.
When I run the test code (python inference.py --question_type Clarification), I get the following error.
How should I handle it?

elasticsearch.exceptions.NotFoundError: NotFoundError(404, 'index_not_found_exception', 'no such index', csqa_wikidata, index_or_alias)

Hi, I have tried to run the whole code base by python=3.7 on Ubuntu 20.04 with cuda=117 and RTX 3090 in anaconda env.
I have go through many library update error especially in the torch-geometric.
I finally successful run all the preprocess step and start to train the model. However I encounter
geometric runtimeError: philox_cuda_state for an unexpected CUDA generator used during capture.
which related to issue and it seems suggest me to update the fix version of torch-geometric you put into the source code, which would ruin all the building by dependency issue. Do you have any idea of this? or do I need to offer more information?

torch1.6 report this error
OSError: /home/admin/anaconda3/lib/python3.7/site-packages/torch_sparse/_convert.so: undefined symbol: _ZN6caffe28TypeMeta21_typeMetaDataInstanceIdEEPKNS_6detail12TypeMetaDataEv

how to solve it, thanks

Hello, thanks for your sharing and contributions.
When I tried to reproduce the code, I encountered an error which said that “ROOT_PATH is not defined” in prepare_data.py. I tried to specify it as the dataset address, but the error did not change.
What does ROOT_PATH.parent mean ？And how to specify the value of ROOT_PATH ?

Hi,
When question type is "Clarification", you concat the next context tokens after current context in dataset._prepare_data and dataset.get_inference_data function, and treat the next system golden action as the currenct golden action, why are you doing this ? I think it will reduce the question number with "Clarification" type, and lead to a unfair compare with others.

Waiting for your responds.
Best,