LangChain4j

How it is done in python: https://github.com/langchain-ai/langchain/blob/9b3962fc2521ec0d6ef2ea7c0a40b9c32977671a/libs/experimental/langchain_experimental/llms/ollama_functions.py

@rh-id `OllamaChatModelIT` is just a bad test, I will improve it

Hi @imbaTapac at the moment only BERT-based embedding models are supported (it is mentioned in the javadoc of OnnxEmbeddingModel)

@imbaTapac I will look into it at some point but don't know when exactly, will try next week.

@sraeuchle sorry did not had time yet, but planning to look into it next week. Please let me know which exactly models do you plan to use, links to huggingface...

@sraeuchle removing `token_type_ids` should be simple, we can add a conditional logic there. Feel free to open a PR!

Ok ok I hear you! Will plan to do that next week.

@sraeuchle please let me know which parts are too strict so that I can relax them. In the past it was definitely the case, but now I am trying to...

@sraeuchle thanks for the feedback! You are doing everyting right, the problem is on our side, I will work to address these issues soon.

These issues should be addressed by https://github.com/langchain4j/langchain4j-embeddings/commit/0878f6e9e3f69c3cb5ffd85fbd6731e7fe0abd8d (which is WIP but probably will be released soon), please review if you have time 🙏