bia-bob
bia-bob copied to clipboard
haesleinhuepf
BIA Bob is a Jupyter+LLM-based assistant for interacting with image data and for working on Bio-image Analysis tasks.
I just noticed, that we don't need to import `bob` explicitly because [this line](https://github.com/haesleinhuepf/bia-bob/blob/9b8db0dbbfade44aac969f2c77dce1d170895743/src/bia_bob/_machinery.py#L37) registers it in the jupyter backend. It also works like this:  I'm not sure how...
``` Chat history: user Load the blobs.tif image file and show it. assistant from skimage.io import imread import stackview image = imread("blobs.tif") stackview.insight(image) user Load the blobs.tif image file and...
@haesleinhuepf I thought it could be useful to tell bob to **use functions** whenever possible such that the code becomes more reusable both within the notebook and also for an...
Pricing
https://openai.com/pricing gpt-3.5 is 20x cheaper than gpt-4.0, quite a difference!
I started exploring how fine-tuning a model would work in [this notebook](https://github.com/haesleinhuepf/BioImageAnalysisNotebooks/blob/79f9ddf3df83af9ad3f18f4bb2f1d02dd6cc4e4a/docs/07_prompt_engineering/80_llm_fine_tuning.ipynb). I don't think the impact is huge and I think we need a lot more training examples (so...
We could use a cheap LLM to create a summary of the previous chat to save input tokens for the more expensive actual LLM that we may use for the...
when giving this to users we should add a couple of example notebooks showing some (non-trivial) image analysis tasks.
Pinecone has been mentioned in a couple of chats that I read: https://archive.pinecone.io/learn/vector-database/ For instance, instead of retraining a model, one could try to put image.sc into such a database....
This is WIP Hardware requirements and loading time are quite huge. Hence, this may never get merged.
Metadata
Owner
Metadata
BIA Bob is a Jupyter+LLM-based assistant for interacting with image data and for working on Bio-image Analysis tasks.