# Dr.Sparse SpGEMM training pool (607 matrices)

The complete RL / selector training pool of Dr.Sparse (branch v2): 607 SuiteSparse matrices in the
harness `.bin` layout (int32 rows, cols, nnz; int32 row_ptr; int32 col_ind; float32 values; float32 x),
laid out as level1_small/ (91), level2_medium/ (273), level3_large/ (198), level4_huge/ (45).
Every matrix has a cuSPARSE SpGEMM reference (C = A*A, or A*A^T when rectangular) on an H200;
the 27 that do not are listed in spgemm_excluded.txt and are not included. None of the OTF-81 test
matrices (level_sparse/otf_test_set.csv in the repo) is included. train.jsonl is the per-matrix
prompt file the RL launcher reads; manifest.csv lists rows, cols, nnz and level.

    python level_sparse/extend_pool.py --hf-download KinGeorge/Dr.Sparse-RL-train-607 --out /big/disk/train607 --link-into level_sparse