Akshita Bhagia

@philz0918 There are a couple of things which can cause the list of instances to be empty. One is that you need to specify the path to the folder containing...

@ashwin153 Yes, I should have mentioned this in the original issue. While it's possible to add `args` and `kwargs` to avoid the error, the underlying problem of having to change...

@Alexixu Possibly your usecase be solved by handling empty instances in the (custom) `DatasetReader` you're using. If not, please share more details on what dataset reader you are running this...

@vikigenius I wrote a small test script, but cannot reproduce the latency that you're observing. ``` import torch from allennlp.training.metrics import ROUGE from allennlp.data.token_indexers.pretrained_transformer_indexer import PretrainedTransformerIndexer token_indexer = PretrainedTransformerIndexer("facebook/bart-base", namespace="tokens")...

@vikigenius I can reproduce the above. I'll look into it.

@Boltzmachine Currently, you have set `CACHEABLE = False` for `Share` step. If you set it to `True`, it will cache the output, and the step will be run only once.

Thanks for looking into this! @martindurant I ensured that the bucket was empty before each run, so I don't think it's an issue with the behavior being different the second...

Also ran into a similar issue. > maybe add an API to the workspace that allows you to delete a cache entry, and then expose it in the CLI? This...

This needs to happen in catwalk, rather than here.

Huggingface recommended the convention to be OLMoXYZ for the naming of the model. Also, I will recommend not making changes to the model on the hub unless absolutely necessary, as...