fix(pinecone_rm): refactored to use cloud_embed and fix pinecone init #1027
Conversation
| ERRORS, | ||
| max_time=15, | ||
| ) | ||
| def get_embeddings(self, queries: List[str]) -> List[List[float]]: |
There was a problem hiding this comment.
could the caching functionality included in this PR be used here?
I realize we probably need an abstracted Retriever class to handle this but for now, feel free to check if this suffices
| { | ||
| "data": { | ||
| "text/plain": [ | ||
| "[upserted_count: 100,\n", |
There was a problem hiding this comment.
maybe better to remove the output for this cell :)
| "name": "stdout", | ||
| "output_type": "stream", | ||
| "text": [ | ||
| "Bootstrapped 0 full traces after 20 examples in round 0.\n" |
There was a problem hiding this comment.
it seems like the bootstrapping in this case couldn't find any valid examples. I'm curious if it is because of the retriever and what passages are being retrieved - worth investigating if it is due to the changes in this PR. We typically expect this example to work with most LMs/RMs but it could just be these 20 were not valid outputs. Just worth checking IMO before merging this example if that is the case
|
Hi @csellis , thanks for the PR! left a few comments to be addressed before merging. Also could you run |
|
Hi @csellis , I updated the ruff fix but accidentally merged this. Could you reopen this with the updated changes as per the comments? Thanks. |
2 participants
source is from https://github.com/stanfordnlp/dspy/pull/342/files