Closed tstadel closed 2 years ago
Hey @tstadel thanks for the contribution. A better DPR evaluation is really needed to steer the training process.
We would take your code and make some adjustments so we can use the evaluation in our Evaluator object. See #716 We will work on it in the coming sprint and will leave this PR open for now. Hope that is fine? Thanks again for this important feature.
Sure, you're welcome!
This issue has been automatically marked as stale because it has not had recent activity. It will be closed in 21 days if no further activity occurs.
This issue has been automatically marked as stale because it has not had recent activity. It will be closed in 21 days if no further activity occurs.
dpr_evaluation.py
script contains average rank evaluation over the full dpr dataset. In order to make this work some minor changes had to be made toTextSimilarityHead
andBaseBiAdaptiveModel
.