Issue with yml file - fixed

[saptarshi059]

Hi,

Thanks for your code. However, the YML file you provided had a TON of broken dependencies. I've uploaded the one that finally worked for me. The issue was that your YML file had a lot of locked system hashes. Also, pip install vllm, was trying to build VLLM, which is never a good strategy. So, it should be downloaded from source (pre-built) [pip install vllm --extra-index-url https://download.pytorch.org/whl/cu121]

I hope this helps with reproducibility. I am testing your code on a few things & will provide feedback if it breaks again.

GraphAnchor_Final.yml

[saptarshi059]

Also, since you are using vLLM with Qwen2.5-7B, it expects this in the config - https://huggingface.co/Qwen/Qwen2.5-7B-Instruct#processing-long-texts

So, it should be mentioned that the config should be updated before using the code.

[Mignonmy]

Hi, Thanks a lot for the effort you put into fixing the YML and sharing the working version. I really appreciate it!

Good point about the vLLM install as well, I’ll update the repo to avoid build issues.

Regarding the Qwen2.5-7B long-text config: in our case, inputs are usually below 8192 tokens, so we haven’t needed it, but I agree it’s worth mentioning for others. I’ll add a note in the docs.

Thanks again, and feel free to share more feedback if anything else comes up!