Hugging Face's logo
Sachin Dohdiya's picture

Sachin Dohdiya

sachin7x
4
·

AI & ML interests

None yet

Recent Activity

liked a dataset 12 days ago
MoreThought/Fable-5.1-Max-Reasoning-Filtered-10000x
liked a model 12 days ago
abenzerps/Qwen-Image-2.1-Uncensored-GGUF
reacted to Ryenhails's post with 🔥 12 days ago
🚀 NanoVDR goes multi-vector: meet ColNanoVDR! Multi-vector VLM retrievers lead visual document retrieval, but every search runs a multi-billion-parameter query encoder. We distill that encoder into a 149M text-only student that queries the teacher's existing page index directly. No re-indexing, and no pages during training. 🧠 How: OTW (Optimal Transport with Learned Weights) aligns the student's query tokens with the teacher's, even though the two tokenize differently (e.g. 17 vs 29 tokens). We prove the alignment cost bounds the MaxSim score gap on every page, so training only needs cached teacher query tokens. 📊 Five teachers → five 149M students, ViDoRe v3 NDCG@5: - ColVec1.1-8b: 62.6 → 60.1 (96.0%) - ColVec1.1-4b: 61.6 → 59.1 (95.8%) - Vultron-4.5B: 61.0 → 58.3 (95.5%) - ColQwen3.5-4.5B: 58.7 → 55.1 (93.8%) - Tomoro-ColQwen3-8B: 59.0 → 54.9 (93.0%) ⚡ 26× faster query encoding on a single CPU thread (87 ms vs 2.3 s) 💾 Matches score distillation while reading 12.6× less cached teacher data 📄 Paper: https://huggingface.co/papers/2609.34899 🤗 Checkpoints (all five students): https://huggingface.co/nanovdr 💻 Code: https://github.com/Ryenhails/NanoVDR 🧩 Single-vector predecessor, NanoVDR: https://arxiv.org/abs/2603.12824 If you already serve one of these teachers, swap in the matching student and keep your index as is. Feedback and upvotes welcome! 🙌
View all activity

Organizations

None yet