Eren Irmak's picture

Eren Irmak PRO

martineden

AI & ML interests

NLP, computational psychiatry, computational neuroscience

Recent Activity

Organizations

Yeditepe NLP Lab's profile picture Hugging Face Discord Community's profile picture

martineden's activity

upvoted an article 15 days ago
view article
Article

nanoVLM: The simplest repository to train your VLM in pure PyTorch

By ariG23498 and 6 others
136
upvoted an article 16 days ago
view article
Article

Open-source DeepResearch – Freeing our search agents

By m-ric and 4 others
1.25k
reacted to AdinaY's post with 👍 about 2 months ago
view post
Post
1839
MAYE🎈a from-scratch RL framework for Vision Language Models, released by GAIR - an active research group from the Chinese community.

✨Minimal & transparent pipeline with standard tools
✨Standardized eval to track training & reflection
✨Open Code & Dataset

Code:
https://github.com/GAIR-NLP/MAYE?tab=readme-ov-file
Dataset:
ManTle/MAYE
Paper:
Rethinking RL Scaling for Vision Language Models: A Transparent, From-Scratch Framework and Comprehensive Evaluation Scheme (2504.02587)
  • 1 reply
·
replied to AdinaY's post about 2 months ago
view reply

I hope this will be integrated into trl soon..

view reply

Thank you for the clarification :huggingface: btw neat tutorial, thanks for sharing

New activity in openfree/Llama-4-Scout-17B-Research about 2 months ago

Update README.md

#1 opened about 2 months ago by
martineden
view reply

At least figure out something on your own. Spoon feeding does not last long.

it's not spoon feeding. it's ambiguity. where did you import Tool to define VisualRAGTool class ? is it smolagents, is it langchain? is it pydantic-ai ? how can one resolve the ambiguity ? you may argue that the tool is available already and can be directly loaded, but I need to implement some tweaks and different setups, so I need unambigious details.