fangguo1

Results 8 comments of fangguo1

the directory "data/preprocessed" needs to be manually created.

Hi, I met the same issue, do you have a solution right now? Thanks!

the directory "data/preprocessed" needs to be manually created.

Did you finally make the model run successfully?

the effectiveness of the trained sledge-med.p model

In your example ( https://colab.research.google.com/drive/1t5UdW2Jebue1php888ldDll6yG5jQXQQ?usp=sharing ), why do you ignore the pooler layer? And it seems in your toy example, the logits are from the token instead of from the...

the effectiveness of the trained sledge-med.p model

Thanks for getting back so quickly. I tried some arbitrary text segments and calculate its relevant score with the query 'Is Hydroxycholoroquine effective?', the result is weird. For example, the...

the effectiveness of the trained sledge-med.p model

Dear Sean, thanks for your reply and your patience. Yes, I would like to use the example using PyTerrier integration.

the effectiveness of the trained sledge-med.p model

Actually, I am not studying typical information retrieval but got inspired by your SledgeZ paper and plan to use a similar zero-shot learning idea on the data in my domain...

ValueError: Too many values to unpack (expected 2) in collection_indexer.py, line 224, _concatenate_and_split_sample

Hi, I got the same error, did you figure out why? Thanks

monoT5 fine-tuning process

@vjeronymo2 Thanks for making this finetune_monoT5.py for reference. However, in this file, since you are using the trainer class, keys in the 'dataset_train' should be the same as the variable...