-
I'm encountering an error while trying to use the WebsiteSearchTool with Gemini Pro as the LLM. The error message is:
"Embedding dimension 768 does not match collection dimensionality 1536. This is…
-
### What happened?
Unable to test this in my organization, as we do not use hugging face cache folders for models, models are downloaded via git, scanned, then allowed for usage. I see some attempt t…
-
Hi,
I'm currently working in a system where I score the document based on a calculated score from two separate retrievals, each one with it's own sbert model (a linear sum of two cosine similarities)…
-
I was flowing a tutorial on YouTube https://www.youtube.com/watch?v=nXpBlC6OBw4&t=840s and I did everything that was lined out, but when i ran the one-click training I got the error (RuntimeError: use…
-
Does this project support the training and inference of multi-modal retrieval models, such as Phi-3-vision? I'd like to reproduce the experiments in paper https://arxiv.org/abs/2406.11251 based on thi…
-
Embedded using the API
Significantly underperforms vs other models
In most of the cases, each embedding is a full text of the Supreme Court decision
Indexed with hnsw.
Should I use a different…
-
Thanks for the impressive work. I tested some videos in qvhighlights val split and follow the prompt: "Find the video segment that corresponds to the given textual query '{}' and determine its start a…
-
BGE Landmark Embedding: A Chunking-Free Embedding Method For Retrieval Augmented Long-Context Large Language Models
-
正在写作中:https://www.oreilly.com/library/view/designing-data-intensive-applications/9781098119058/
翻译计划于 v2 分支进行。
- [ ] Chapter 1: Tradeoffs in Data Systems Architecture (available)
- [ ] Chapter …
-
### Model description
[Monarch Mixer BERT](https://hazyresearch.stanford.edu/blog/2023-07-25-m2-bert) is sub-quadratic in sequence length. This has enabled the development of [32k-context retrieval m…