Prithiv Sakthi's picture

Prithiv Sakthi PRO

prithivMLmods

·

https://huggingface.co/strangerzonehf

AI & ML interests

computer vision, multimodality, realism engine adapters @starngerzonehf

Recent Activity

reacted to mkurman's post with ❤️ 7 minutes ago

Blurred-Thoughts Supervised Fine-Tuning (BT-SFT) 🤖 Can we teach a model to think completely on its own without reinforcement learning? Actually, yes. We can do straightforward supervised fine-tuning using a relatively simple trick: blurring a part of CoT thoughts. But why is this effective? We observed that various models differ in their thinking processes, and fine-tuning one model on another model’s thoughts (CoT) can sometimes be inefficient—often resulting in the model simply memorizing reasoning rather than learning how to actually think. I discovered that this process can still be efficient if we clearly indicate when the model should start and stop thinking and uncover only a part of CoT and the expected answer, blurring the other part of CoT. This approach allows the model to learn only a portion of the thought process while still arriving at an expected answer. To demonstrate this, you can watch my experimental BT-SFT on meditsolutions/Llama-3.2-SUN-2.5B-chat model, which was fine-tuned on 151 million tokens from the Magpie-Align/Magpie-Reasoning-V2-250K-CoT-Deepseek-R1-Llama-70B dataset. Enjoy! 🚀 PS. If you were curious enough to read this, leave me a comment. It's always nice to chat with open-minded and intelligent ppl.

new activity 33 minutes ago

onnx-community/convert-to-onnx:Upload ONNX to Your Current Model Repo 📁

updated a model about 1 hour ago

prithivMLmods/FastThink-0.5B-Tiny

View all activity

Articles

Why PyThagorean? 🐍

N-Queens Problem Based Monte Carlo Algorithm

Fine-tune SmolLM's on custom synthetic data

GRID-6X : Layout for Seamless Image Assembly

Flux1.1 [pro] Ultra : Endpoint by BFL ⛵

Create Dynamic Typed Videos with 'Type Byte🐧'

Unlocking Creativity with Text-to-Image Generation: Exploring LoRA Models and Styles [Generative Vision]

Organizations

prithivMLmods's activity

upvoted an article about 17 hours ago

Article

KV Caching Explained: Optimizing Transformer Inference Efficiency

By

•

1 day ago

• 19

upvoted a collection about 17 hours ago

Bellatrix Tiny 🤏🏻

1.5B, 1B, 0.5B param's • 5 items • Updated about 17 hours ago • 5

upvoted an article 2 days ago

Article

Welcome to Inference Providers on the Hub 🔥

3 days ago

• 171

upvoted a collection 4 days ago

Qwen2.5-VL

Vision-language model series based on Qwen2.5 • 3 items • Updated 4 days ago • 287

upvoted 2 collections 5 days ago

Reasoning Saturations🫙

7B, 14B [ Saturation ] • 6 items • Updated about 17 hours ago • 7

Opus Elite 💭

Model contention for gradual increments in reasoning-specific model evaluations. • 5 items • Updated 2 days ago • 7

upvoted an article 6 days ago

Article

Failure Modes of OpenAI Operator

By

•

7 days ago

• 3

upvoted 2 articles 7 days ago

Article

Explore, Curate and Vector Search Any Hugging Face Dataset with Nomic Atlas

By

•

7 days ago

• 29

Article

Fine-tune ModernBERT for text classification using synthetic data

By

•

Dec 30, 2024

• 28

upvoted 2 articles 8 days ago

Article

Mastering Long Contexts in LLMs with KVPress

By

•

8 days ago

• 58

Article

Fine-tune ModernBERT for RAG with Synthetic Data

By

•

10 days ago

• 28

upvoted 2 collections 8 days ago

Calcium-Opus 💬

Opus for chat • 3 items • Updated 7 days ago • 6

Phi-4 Custom FT

phi 4 custom ft [ qwq, lcot, math ] • 7 items • Updated 6 days ago • 8

upvoted a collection 9 days ago

Express 🚅

Express Tiny LLM's • 6 items • Updated 6 days ago • 7

upvoted 2 collections 10 days ago

QwQ Reasoning 🐤

conversational, instruct, 14b Convos < 350 in OpenLLM Evals • 5 items • Updated 8 days ago • 7

Q-Series Sketch

Q(n) • 5 items • Updated 10 days ago • 7

upvoted a collection 11 days ago

Doge-SLM

Doge family of small language models • 7 items • Updated 5 days ago • 5

upvoted an article 11 days ago

Article

Why PyThagorean? 🐍

By

•

11 days ago

• 7

upvoted 2 collections 11 days ago

PyThagorean 🐍

Solve Math with Python • 3 items • Updated 11 days ago • 7

LwQ -Llama 🦙

Llama with Questions for reasoning / long chain of thought • 3 items • Updated 11 days ago • 7