OSS contributions

I’ve collaborated in several 🤗 Hugging Face repos/projects. Here, you can find the full list of current work, closed PRs, and opened PRs in their repos.

Highlighted projects

• Developed 10 recipes focused on cutting-edge techniques, including multimodal models (VLMs, RAGs), computer vision, and agent-based systems, featuring technologies such as ColPali/ColQwen and Qwen2-VL.

  • Scaling Test-Time Compute for Longer Thinking in LLMs - PR
  • Fine-tuning SmolVLM using direct preference optimization (DPO) with TRL on a consumer GPU - PR
  • Fine-tuning SmolVLM with TRL on a consumer GPU - PR
  • Smol Multimodal RAG: Building with ColSmolVLM and SmolVLM on Colab’s Free-Tier GPU - PR
  • Multimodal RAG with ColQwen2, Reranker, and Quantized VLMs on Consumer GPUs - PR
  • Multi-agent RAG System 🤖🤝🤖 - PR
  • Fine-Tuning a Vision Language Model (Qwen2-VL-7B) with the Hugging Face Ecosystem (TRL) - PR
  • Multimodal Retrieval-Augmented Generation (RAG) with Document Retrieval (ColPali) and Vision Language Models (VLMs) — PR
  • Fine-Tuning a Semantic Segmentation Model on a Custom Dataset and Usage via the Inference API — PR
  • Fine-Tuning Object Detection Model on a Custom Dataset 🖼, Deployment in Spaces, and Gradio API Integration — PR
• Created 2 recipes for Llama 3.2, covering its deployment options with TGI and how to implement tool calling.
  • Tool Calling using Llama 3.2 🔧 - PR
  • 🦙 Using the Text Generation Inference (TGI) with Llama Models — PR
• Currently exploring advanced VLM models and their integration with TRL, with contributions already made in this area.
  • 👁️ Added SFT support for SmolVLM models via standalone script sft_vlm_smol_vlm.py — PR

I am one of the top contributors to the following repositories:

• Cookbook
• Huggingface-llama-recipes

Current work (Nov 24)

Transformers

• Support for Multiple Datasets and Domain-Specific Loss Calculation in Trainer #30725 - PR

Cookbook

TRL

• [SFT VLM] Add support for Molmo models #2136 — PR

Closed PRs

TRL

• 👁️ Added SFT support for SmolVLM models via standalone script sft_vlm_smol_vlm.py #2409 — https://github.com/huggingface/trl/pull/2409
• Updated ScriptArguments warning messages #2230 - https://github.com/huggingface/trl/pull/2230

Cookbook

• Added Multi-agent RAG system recipe 🧑‍🍳️ #238 - https://github.com/huggingface/cookbook/pull/238
• Updated Agentic RAG recipe #232 - https://github.com/huggingface/cookbook/pull/232
• Added Recipe for Fine-Tuning a Vision Language Model (Qwen2-VL-7B) with the Hugging Face Ecosystem (TRL) #225 - https://github.com/huggingface/cookbook/pull/225
• Added Cookbook for Multimodal Retrieval-Augmented Generation (RAG) with Document Retrieval (ColPali) and Vision Language Models (VLMs) #210 — https://github.com/huggingface/cookbook/pull/210
• Updated broken image link in Fine-Tuning Object Detection on a Custom Dataset… notebook #204 — https://github.com/huggingface/cookbook/pull/204
• Add Notebook for Fine-Tuning Semantic Segmentation Models and Using the Inference API #201 — https://github.com/huggingface/cookbook/pull/201
• Added missing HF login cell in Agent for text-to-SQL with automatic error correction notebook #198 — https://github.com/huggingface/cookbook/pull/198
• Updated introduction and installation instructions #196 — https://github.com/huggingface/cookbook/pull/196
• Have several agents collaborate in a multi-agent hierarchy 🤖🤝🤖 notebook improvements #194 — https://github.com/huggingface/cookbook/pull/194
• Removed duplicated Open in Colab button in Semantic reranking with Elasticsearch and Hugging Face #192 — https://github.com/huggingface/cookbook/pull/192
• Object Detection Fine tuning cookbook with Deployment in Spaces, and Gradio API Integration #189 — https://github.com/huggingface/cookbook/pull/189
• Updated code in the notebooks in Chinese to match English versions #178 — https://github.com/huggingface/cookbook/pull/178
• Paragraph refined in Build RAG with Hugging Face and Milvus #176 — https://github.com/huggingface/cookbook/pull/176
• Indentation update in RAG backed by SQL and Jina Reranker v2 cookbook #174 — https://github.com/huggingface/cookbook/pull/174
• Corrected malformed URL #172 — https://github.com/huggingface/cookbook/pull/172
• Typos fixed in Using LLM-as-a-judge 🧑‍⚖️ for an automated and versatile evaluation #170 — https://github.com/huggingface/cookbook/pull/170
• Migrating from OpenAI to Open LLMs Using TGI’s Messages API notebook updated #167 — https://github.com/huggingface/cookbook/pull/167
• Updated table rendering working in Colab in “How to use Inference Endpoints to Embed Documents” #164 — https://github.com/huggingface/cookbook/pull/164
• Added HF login, and OpenAI and Anthropic packages installation #149 — https://github.com/huggingface/cookbook/pull/149
• Fix minor issues in agents notebook when run using Colab #147 — https://github.com/huggingface/cookbook/pull/147
• Fixed small typo in agents notebook #143 — https://github.com/huggingface/cookbook/pull/143
• Fix Issues with Colab Notebook in ‘Annotate text data using Active Learning with Cleanlab’ #139 — https://github.com/huggingface/cookbook/pull/139
• Restored broken link to FAQ #137 — https://github.com/huggingface/cookbook/pull/137
• Patch to enable notebook functionality with a workaround #132 — https://github.com/huggingface/cookbook/pull/132
• Restored broken link in Fine-tuning a Code LLM on Custom Code on a single GPU notebook #130 — https://github.com/huggingface/cookbook/pull/130
• Fixed typo in RAG with Gemma+ES+HF cookbook #128 — https://github.com/huggingface/cookbook/pull/128
• Solved small issues when running RAG evaluation notebook in Colab #126 — https://github.com/huggingface/cookbook/pull/126
• Updated advanced RAG notebook dependencies #121 — https://github.com/huggingface/cookbook/pull/121
• Added langchain-community pip install #116 — https://github.com/huggingface/cookbook/pull/116

Huggingface-llama-recipes

• Tool calling recipe #84 - https://github.com/huggingface/huggingface-llama-recipes/pull/84
• Adding Using the Text Generation Inference (TGI) with Llama Models #69 — https://github.com/huggingface/huggingface-llama-recipes/pull/69
• Small punctuation update in Run Llama with AWQ notebook #12 — https://github.com/huggingface/huggingface-llama-recipes/pull/12

Setfit

• Renamed evaluation_strategy to eval_strategy #538 — https://github.com/huggingface/setfit/pull/538

Blog

• Updated broken notebook link in Llama 3.2’s blog — https://github.com/huggingface/blog/pull/2371
• “The 5 Most Under-Rated Tools on Hugging Face” post improvements #2320 — https://github.com/huggingface/blog/pull/2320
• Typo fixed in Open-source LLMs as LangChain Agents blog #2288 — https://github.com/huggingface/blog/pull/2288

Computer-vision-course

• Updated broken link in Unit 3 - Swin transformer #346 — https://github.com/johko/computer-vision-course/pull/346
• Embedded space URL corrected #339 — https://github.com/johko/computer-vision-course/pull/339
• Removed sentence pointing to missing video #332 — https://github.com/johko/computer-vision-course/pull/332
• Object detection fine-tuning code issues #331 — https://github.com/johko/computer-vision-course/pull/331
• Updated Math description to correct rendering #316 — https://github.com/johko/computer-vision-course/pull/316
• Added credits as a reviewer #310 — https://github.com/johko/computer-vision-course/pull/310
• Added missing quotes to sentence #276 — https://github.com/johko/computer-vision-course/pull/276
• Small typo fixed #266 — https://github.com/johko/computer-vision-course/pull/266
• Italic rendering problem updated #265 — https://github.com/johko/computer-vision-course/pull/265
• Reviewed general punctuation and fixed some broken links #251 — https://github.com/johko/computer-vision-course/pull/251
• Removed audio from Ethics and Biases unit #247 — https://github.com/johko/computer-vision-course/pull/247

Transformers

• Agents, supercharged - Multi-agents, External tools, and more docs typo fixed #33478 https://github.com/huggingface/transformers/pull/33478
• Corrected Agents and tools documentation links typos #33471 — https://github.com/huggingface/transformers/pull/33471
• Fix incomplete sentence in Zero-shot object detection documentation #33430 — https://github.com/huggingface/transformers/pull/33430
• Fixed typo repeated word in DETR docs #33250 — https://github.com/huggingface/transformers/pull/33250

Smol-vision

• Updated installation using pip in ColPali_+_Qwen2_VL.ipynb #12 — https://github.com/merveenoyan/smol-vision/pull/12

Open PRs

smolllmm

• Updated FT notebook with small note about CUDA_VISIBLE_DEVICES #21 - https://github.com/huggingface/smollm/pull/21

Cookbook

• Updated multiagent notebook with HF login and solving model naming inconsistencies #234 - https://github.com/huggingface/cookbook/pull/234
• Added HF login cell to Build an agent with tool-calling superpowers 🦸 using Transformers Agents recipe #230 — https://github.com/huggingface/cookbook/pull/230

Evaluate

• Typo fixed in Types of Evaluations in 🤗 Evaluate documentation #640 — https://github.com/huggingface/evaluate/pull/640

Notebooks

• Updated Object detection task notebooks #515 — https://github.com/huggingface/notebooks/pull/515
• Removed duplicated login cell in NLP course notebook #511 — https://github.com/huggingface/notebooks/pull/511

NLP Course

• Updated Github rate limit URL #740 — https://github.com/huggingface/course/pull/740
• Updated number of models in HF Hub #738 — https://github.com/huggingface/course/pull/738
• Updated files structure in 4. Sharing models and tokenizers - Sharing pretrained models #736 — https://github.com/huggingface/course/pull/736
• Updated authentication instructions: replace username/password with token #733 — https://github.com/huggingface/course/pull/733
• Updated dataset features ClassLabel inconsistent output #731 — https://github.com/huggingface/course/pull/731
• Updated .bin to .safetensors in 2. Using 🤗 Transformers - Models - Saving methods #729 — https://github.com/huggingface/course/pull/729
• Updated BERT checkpoint link updated #727 — https://github.com/huggingface/course/pull/727
• Typo fixed in Chapter 4 - Building a model card #725 — https://github.com/huggingface/course/pull/725
• Fix typo in title: corrected missing brackets in sequence-to-sequence models #724 — https://github.com/huggingface/course/pull/724
• Corrected link redirection in 7. Main NLP Tasks - Token classification #722 — https://github.com/huggingface/course/pull/722

Audio-transformers-course

• Updated Makefile small typo #189 — https://github.com/huggingface/audio-transformers-course/pull/189

Hug-docs

• Updated index punctuation #1371 — https://github.com/huggingface/hub-docs/pull/1371