Async-GRPO - Open, Fast, and Performant
Written by Aldo Pareja, Mustafa Eyceoz Introducing Async-GRPO With the incr...
LLM Toolkit
sdg_hub
Synthetic data generation pipelines for post-training
Synthetic data generation pipelines for post-training
training
Efficient messages-format SFT library for language models
Efficient messages-format SFT library for language models
async-grpo
Asynchronous GRPO for scalable reinforcement learning
Asynchronous GRPO for scalable reinforcement learning
SQuat
KV cache quantization for scaling inference time
KV cache quantization for scaling inference time
orthogonal-subspace-learning
Adaptive SVD-based continual learning method for LLMs with negligible catastrophic forgetting.
Adaptive SVD-based continual learning method for LLMs with negligible catastrophic forgetting.
probabilistic-inference-scaling
Inference-time scaling with particle filtering.
Inference-time scaling with particle filtering.
Recent posts
View all posts
Sculpting Subspaces: How We Solved Continual Learning in Large Language Models
Authors: Nikhil Shivakumar Nayak, Krishnateja Killamsetty, Ligong Han, Abhish...
Update 3 - On Reasoning vs Inference-time scaling - Lessons on Reproducing R1-like Reasoning in Small LLMs without using DeepSeek-R1-Zero (or its derivatives)
Written by Akash Srivastava, Isha Puri, Kai Xu, Shivchander Sudalairaj, Musta...