TorchEBM Benchmarks
Performance tracking for TorchEBM
Loading benchmark data...
Median Time by Component
Throughput (Samples/sec)
Peak GPU Memory (MB)
Scaling: Time vs Batch Size
About TorchEBM
TorchEBM is a PyTorch library for energy-based modeling, with support for flow and diffusion methods. It provides composable tools for defining energy landscapes, training with various objectives (flow/equilibrium matching, contrastive divergence, score matching), and sampling via MCMC, optimization, or learned continuous-time dynamics (flow-based).