Skip to content
This repository has been archived by the owner on Nov 3, 2023. It is now read-only.

Update Hallucination README.md #3888

Merged
merged 7 commits into from
Aug 4, 2021
Merged

Update Hallucination README.md #3888

merged 7 commits into from
Aug 4, 2021

Conversation

klshuster
Copy link
Contributor

Patch description
Per #3872, we should include the --model-file parameter in the example training commands for the Hallucination project

@jianguoz
Copy link
Contributor

jianguoz commented Aug 4, 2021

Patch description
Per #3872, we should include the --model-file parameter in the example training commands for the Hallucination project

Per #3881 and #3873. In #3881, the model should immediately raise an AssertionError when there are NaN instead of logging errors and/or replacing NaN with other values. Besides, after careful investigation, I found the NaN error is not caused by either --indexer-type exact or --indexer-type compressed. The error comes from num_shards if the passages file format meets requirements and there are no repeated entries. Where if end_idx<n_docs, it will raise NaN error. So it is better to set --num-shards to small values. For example, if --shard-id is 0, then --num-shards is a value < int(len(rows of the passages file)/(n_maximum_passages_you_want_to_retrieve-1)). I had slightly updated the readme also to give more instructions.

Raise AssertionError when `NaN` values appear.
slightly add more instructions to the readme file in RAG
@jianguoz jianguoz changed the title Update Hallucination README.md Update Hallucination README.md and handle NaN error in Retriever Aug 4, 2021
@klshuster
Copy link
Contributor Author

Hi @jianguoz, though I appreciate the changes you have proposed, these seem more suitable for a separate PR (as now I cannot actually review the changes you've made)

@jianguoz
Copy link
Contributor

jianguoz commented Aug 4, 2021

Hi @jianguoz, though I appreciate the changes you have proposed, these seem more suitable for a separate PR (as now I cannot actually review the changes you've made)

@klshuster Sure! I will do it:).

Copy link
Contributor

@spencerp spencerp left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

lgtm!

@spencerp spencerp changed the title Update Hallucination README.md and handle NaN error in Retriever Update Hallucination README.md Aug 4, 2021
@klshuster klshuster merged commit 9b5f8c6 into master Aug 4, 2021
@klshuster klshuster deleted the klshuster-patch-1 branch August 4, 2021 18:15
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Projects
None yet
Development

Successfully merging this pull request may close these issues.

4 participants