[draft] new prompt templates - nearly 100% completition: ollama + qwen.2-5-instruct:32b (14b/7b feasible too) #327
Replies: 5 comments 7 replies
|
Hi, Best Regards |
|
I'm still sticking with it - haven't added that many more documents lately. So I didn't get around implementing any changes yet ... I have setup an automation for adding the ocr tags for new documents however ... that`s also how I got around re-doing everything again and again for testing purposes. Feel free to try and report back! I guess we might need better some code changes for better results. Yesterday I`ve been looking into letta ... not sure if that's overkill but I'll see. |
|
Short update since I'm currently developing an LMM based Search Engine that makes use of these models too (https://github.com/thiscantbeserious/llm-search-agent)
Albeit it does seem to have a much simpler language level, this should be fine for document titles and such. Will keep you updated with my observations since I'm planning to implement a testing pipeline to benchmark the accuracy and reliability there against my prompts. |
|
Any updates for a better model for german documents? |
|
Short Update: I'm testing the new ollama version with Vulkan support on an UGreen DXP4800 (Intel N100) now with 32GB Ram ... I recon I will try going to opt for |
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
From my testings I figured that
deepseek-r1wasn't able to properly follow the instructions especially in regards to parsing the date. That's why I tried a few different models and landed atqwen2.5:32b-instruct(7b and 14b work too, just slightly more errors).This produces the most reliable results for me - including documents that weren't properly OCR`ed given my instructions.
I ran multiple models against the full 178 documents multiple times and different document-types including hand-writing.
Key take-away is that we might need a few improvements on the code-layer to improve other model results. For example the first thing I'd add is a simple date parser that extracts the last date that matches YYYY-MM-DD to solve the issue with slight off responses with additional noise. However models like deepseek-r1 would highly improve if we are able to use multiple prompts to fine-tune results, so that`s another possible improvement I do see on the code layer. And most importantly we should allow different models for each field - say you know that an instruct model is far better at following instructions like date parsing, but you want a more refined model like deepseek-r1 to work on the title.
That's the reasoning behind my choice and some outlook of possible PR`s / Improvements I might see ...
I noticed that the most important change was moving the Content at the top of the template for the instructions to not be overwritten under any circumstance.
paperless-gpt
TOKEN_LIMITset at:3000Ollama default Context Length via
OLLAMA_CONTEXT_LENGTH:8096-4048should be fine too ...Language of Documents (and System):
German
Documents tested:
179 so far and counting
Success rate:
Nearly 100% - Quality I would rate at 70-80% of the results being ok, even on really sub-par OCRs ...
I created and refined the prompts with the models itself to make sure its in a format its working best against - last but not least however I manually fine-tuned them and tested them again until I was somewhat satisfied ...
You can find the most recent version of the templates in my Git-Repo:
https://github.com/thiscantbeserious/paperless-gpt-prompts
All reactions