-
In the replication of the MS MARCO passage ranking experiment, we retrieve passages by running
```
python ./src/main/python/msmarco/retrieve.py --hits 1000 --threads 1 \
--index indexes/msmarco-pa…
-
I was trying to use custom data for passage ranking by creating a task with data_format 'PremiseAndMultiHypothesis'. While running train.py with RankCeCriterion, I'm getting the following error:
```
…
-
i have some questions about Passage ranking dataset:
a) there are approximately 40W different querys in _triples.train.full.tsv.gz_, are those all in _qrels.train.tsv_ & _qrels.dev.tsv_ files?
b) fu…
-
Hi, I have read the README.md several times but still don't know how to get .trec.with_json. May be I missed something, Can you tell me some details about it, thanks.
-
In issue #1301 @westnordost suggested to open a new issue regarding quest profiles, i.e. different quest selections and priorities to select depending on the current situation.
Sorry if this issue is…
-
Hi there, I notice that, in SuperGlue, t5 handle ReCoRD by concatenating all candidate answers after the question and before the passage, and let the model to generate the correct answer. But when the…
-
Hi,
I am playing with the dataset you have generated (the one behind this link https://drive.google.com/open?id=1-LZcCSVwejkdMg_9rnteUi2kHBZT-NP_).
I have trained a model based on Pytorch levera…
-
Follow up to #1167
I ran this:
```
$ python src/main/python/run_es_regression.py --regression msmarco-doc --input /tuna1/collections/msmarco/doc
...
2020-05-08 14:34:32,376 INFO - [SUCESS] 0…
-
Hi,
I am trying to reproduce the results on msmarco-passage, and I could not get `train.py` to run. Perhaps the readme is incomplete? What I've done so far:
1. I ran `./generate_file_split.sh` on …
-
I'm new to this project and I want to build index on MS-MARCO documents dataset, for document ranking task.
The readme mentions using following command to start indexing.
```
nohup sh target/ap…