dora-rs
Create a dora-transformers node that can handle multiple model and modality!
In order to facilitate integration of multiple AI models, I think it could be good to have a generic dora-transformers node that can be used to try different AI models.
I think that we can reuse most of the work already made on dora-qwen2.5vl and build on top of it.
![github-actions[bot] avatar](https://avatars.githubusercontent.com/in/15368?v=4 "Published by a pub.dev verified publisher"){kind=small}
Yes still working on the issue. How do i test a node completely through different inputs? I am currently changing models based on environment variables and to test these models i have to use different inputs. Currently i plan to run python scripts separately for testing. is there a more efficient method? @haixuanTao
Hey, I am just curious. what do you mean different inputs? Are u planning to pass image inputs??(for ex: dora-qwen2.5l) to see the output??. I believe as long as your model is up and running and you are following the chat template example or various chat template/ways you should be good to go. Please correct me if i am wrong!!
Can I lend a hand? If it's ok!! multi model modality sounds interesting.
Hello @Mati-ur-rehman-017, you have been unassigned from this issue.
Hey, while developing the dora-llama node I had previously made the node using transformers along with support for multiple models, @haixuanTao can I use that for this transformer node, or is there something else needed in this node?
Yes, you can just use the transformers code and just remove all support for dora-llama-cpp-python making them strictly separated.
Ok going to work on this this weekend
So I have added a minimal testing dataflow in https://github.com/dora-rs/dora/blob/main/tests/llm/phi4.yaml that we can have a look and test very simple prompt and check for the response.
