Skip to main content

Unanswered Questions

5 questions with no upvoted or accepted answers
2 votes
0 answers
68 views

How do I fix the system prompt so that ChatGPT does something meaningful on the web?

My question got closed on SO, I hope this is the appropriate place for it. I'm making an agent framework which enables ChatGPT-4o to interact with the web with (mostly) selenium. Now it does seem to ...
0 votes
0 answers
14 views

How can I accurately measure TOPS during inference of a VLM model (Qwen 2.5 VL 3B Instruct) using `torch.profiler`?

I’m trying to measure TOPS (tera operations per second) during inference of a vision-language model such as Qwen 2.5 VL 3B Instruct. I know we can compute FLOPs and convert to TOPS by dividing by 10¹²,...
0 votes
0 answers
10 views

How to use EcomID model?

The project is trying to create a school-kid-dream-job-visualizer. The LindenIn post that we're trying to achieve: https://www.linkedin.com/posts/futureof-ai_its-a-great-way-to-show-kids-its-...
0 votes
0 answers
16 views

GPT-4o-vision for extraction of complex tables

I am trying to extract complex tables from pdf as markdown using gpt-4o vision preview version. The model does a good job in extracting simple tables but fails to extract complex tables with grouped ...
0 votes
1 answer
75 views

Runing Mistral 7B locally

I used Mistral with Ollama, but IDK how to fine-tuning the model in this way. Therefore I download the model from Hugging Face. I have used this snippet Python: from transformers import ...