Hugging Face's logo
Prithiv Sakthi's picture
Building on HF
606 968 1745

Prithiv Sakthi PRO

prithivMLmods
hugging-science
Realpat's profile picture Peter-Young's profile picture rguillory's profile picture
·
https://linktr.ee/prithivsakthi

AI & ML interests

computer vision, nlp, multimodality - HuggingFace Fellow ML 🤗

Recent Activity

reacted to lbourdois's post with 🔥 26 minutes ago
New blog post! An introduction to a little-known but highly effective model reduction method: 𝗧𝗿𝗶𝗺𝗺𝗶𝗻𝗴✂️ We show how to reduce model size (we went up to 87.24% reduction) while preserving its performance. We applied this technique to 16 different model families across several modalities to illustrate that it works on any architecture (as long as the embedding layer is the last one of the model) and on any modality involving text. From these 16 families, we generated over 𝟱,𝟱𝟬𝟬 𝗺𝗼𝗻𝗼𝗹𝗶𝗻𝗴𝘂𝗮𝗹 𝗺𝗼𝗱𝗲𝗹𝘀 𝗶𝗻 𝟭𝟮𝟰 𝗱𝗶𝗳𝗳𝗲𝗿𝗲𝗻𝘁 𝗹𝗮𝗻𝗴𝘂𝗮𝗴𝗲𝘀 🌍 Key takeaways from our experiments: 1️⃣ Trimming does not require a GPU. Our models were obtained on a CPU. 2️⃣ This method scales up to at least 4B parameters (we did not test beyond that). 3️⃣ Trimmed model is smaller than the original while preserving its performance. If you observe a slight performance drop, just fine-tuned to recover or even surpass the original performance. 4️⃣ For an equivalent compute budget, it is better to trim then fine-tune rather than fine-tuning the original model. Since the model is smaller, you can run more epochs/show more data and get in fine a better model than the original. 5️⃣ Trimming is a competitive alternative to distillation and quantization. E.g. we obtained our alternative to DistilBERT in 9 minutes on CPU vs. 90 hours of GPU for the latter. 6️⃣ Trimming could generate reasoning traces in the language of the trimmed model. This could be an alternative to generating traces in English and then translating them into the desired language. And many other things (such as how much data are needed, the impact of the database used, the order in which it should be done, etc.) are available in the blogpost! Blogpost: https://huggingface.co/blog/lbourdois/introduction-to-trimming Models: https://huggingface.co/spaces/alphaedge-ai/Trimming_models_search
reacted to lbourdois's post with 🤗 26 minutes ago
New blog post! An introduction to a little-known but highly effective model reduction method: 𝗧𝗿𝗶𝗺𝗺𝗶𝗻𝗴✂️ We show how to reduce model size (we went up to 87.24% reduction) while preserving its performance. We applied this technique to 16 different model families across several modalities to illustrate that it works on any architecture (as long as the embedding layer is the last one of the model) and on any modality involving text. From these 16 families, we generated over 𝟱,𝟱𝟬𝟬 𝗺𝗼𝗻𝗼𝗹𝗶𝗻𝗴𝘂𝗮𝗹 𝗺𝗼𝗱𝗲𝗹𝘀 𝗶𝗻 𝟭𝟮𝟰 𝗱𝗶𝗳𝗳𝗲𝗿𝗲𝗻𝘁 𝗹𝗮𝗻𝗴𝘂𝗮𝗴𝗲𝘀 🌍 Key takeaways from our experiments: 1️⃣ Trimming does not require a GPU. Our models were obtained on a CPU. 2️⃣ This method scales up to at least 4B parameters (we did not test beyond that). 3️⃣ Trimmed model is smaller than the original while preserving its performance. If you observe a slight performance drop, just fine-tuned to recover or even surpass the original performance. 4️⃣ For an equivalent compute budget, it is better to trim then fine-tune rather than fine-tuning the original model. Since the model is smaller, you can run more epochs/show more data and get in fine a better model than the original. 5️⃣ Trimming is a competitive alternative to distillation and quantization. E.g. we obtained our alternative to DistilBERT in 9 minutes on CPU vs. 90 hours of GPU for the latter. 6️⃣ Trimming could generate reasoning traces in the language of the trimmed model. This could be an alternative to generating traces in English and then translating them into the desired language. And many other things (such as how much data are needed, the impact of the database used, the order in which it should be done, etc.) are available in the blogpost! Blogpost: https://huggingface.co/blog/lbourdois/introduction-to-trimming Models: https://huggingface.co/spaces/alphaedge-ai/Trimming_models_search
updated a collection 32 minutes ago
MTP Qwen 3.5/3.6 Stable
View all activity

Organizations

OpenGVLab's profile picture Hugging Face Fellows's profile picture DeepGHS's profile picture TorchGeo's profile picture Chinese-Vicuna's profile picture ETH Zürich's profile picture Binghamton University's profile picture Blog-explorers's profile picture OpenSky's profile picture The University of Melbourne's profile picture Johns Hopkins University's profile picture MLX Vision's profile picture LocalLLaMA's profile picture MLX Community's profile picture Agora's profile picture Cohere Labs Community's profile picture M4-ai's profile picture Chinese LLMs on Hugging Face's profile picture ONNX Community's profile picture Hugging Face Discord Community's profile picture ParsBench's profile picture Dataset Tools's profile picture Académie Du Numérique's profile picture Stranger Zone's profile picture open/ acc's profile picture Data Is Better Together Contributor's profile picture None yet's profile picture Taiwan Llama's profile picture Doge Face's profile picture LiteRT Community (FKA TFLite)'s profile picture Stranger Guard's profile picture Bitsandbytes Community's profile picture zjudai's profile picture PowergenAI's profile picture LeRobot Worldwide Hackathon's profile picture Hugging Face Context Course's profile picture Agents-MCP-Hackathon's profile picture Moon AI's profile picture Qualifire's profile picture Hugging Science's profile picture MCP-1st-Birthday's profile picture nanochat students's profile picture Stranger Vision's profile picture Stranger Operations's profile picture Stranger Tools's profile picture Prithivcolab's profile picture Build Small Hackathon's profile picture