dancinlab/clm-v1-ref-pytorch-cuda

Text Generation • Updated 2 days ago

Note Lane-G-ref PyTorch+CUDA baseline (ByteGPT 85.6M, util MEAN 98.85% reference) — NOT forge production (a_train_flame_forge)

dancinlab/clm-v1-ref-pytorch-cuda-7b

Text Generation • Updated 2 days ago • 1

Note Lane-G-ref 7.25B PyTorch-CUDA reference (descent PASS, util 99.18%, bounded N=400, NOT converged, NOT forge)