ThinkingBox Turns Database-State Checks into Agent Training Rewards
Microsoft and Hugging Face have released ThinkingBox-Training on GitHub, including the data-selection, rollout, reward, and Verl integration used in their Qwen3.8-27B experiment.
LLM Training Infrastructure
7 posts
SageMaker AI adds customizable serverless training sessions
SageMaker AI serverless model customization training sessions are generally available, with teams writing their own post-training loops while AWS manages GPU allocation, distributed execution, recovery, and checkpoint storage.
Amazon SageMaker AI adds APIs for custom training loops
Amazon SageMaker AI has introduced training session APIs that let developers define their own training loop while the service operates the GPU infrastructure.
NVIDIA Demonstrates Eight-Node DGX Spark Clusters
Eight DGX Spark nodes combine for 1 TB of unified memory, enough to load the full GLM 5.3 753B checkpoint locally.
Training Qwen3 on AWS Trainium with torchtitan
AWS shows Qwen3 training on one trn2.48xlarge using torchtitan and TorchNeuron, which exposes Trainium through PyTorch’s device interface.
i14 Trains and Serves Laplace 3B on AWS Trainium2
i14’s Laplace 3B uses Gated Linear Attention, keeping its recurrent state at 29.5 MB per sequence regardless of context length.
DeepSpeed v0.19.8 expands AutoTP, AutoEP, and accelerator support
DeepSpeed v0.19.8 adds AutoTP support for vocabulary-parallel LM heads and embeddings, and extends support for tied models and row-parallel output-head training.