From 20d1561a545551393ede3d5ebafcd42f7dc07659 Mon Sep 17 00:00:00 2001 From: Xinqi Li Date: Mon, 16 Mar 2026 07:49:13 +0000 Subject: [PATCH] add scripts for evaluation subset selection --- ...graph_sample_uids_and_op_seq_bucket_id.sql | 22 +++++++++++++++++++ .../select_representive_sample_uids.viba | 11 ++++++++++ 2 files changed, 33 insertions(+) create mode 100644 graph_net/sqlite_util/select_fusible_graph_sample_uids_and_op_seq_bucket_id.sql create mode 100644 graph_net/sqlite_util/select_representive_sample_uids.viba diff --git a/graph_net/sqlite_util/select_fusible_graph_sample_uids_and_op_seq_bucket_id.sql b/graph_net/sqlite_util/select_fusible_graph_sample_uids_and_op_seq_bucket_id.sql new file mode 100644 index 000000000..35f62fe68 --- /dev/null +++ b/graph_net/sqlite_util/select_fusible_graph_sample_uids_and_op_seq_bucket_id.sql @@ -0,0 +1,22 @@ +CREATE TEMP VIEW v_sample_buckets_groups AS +SELECT + g.sample_type, + g.relative_model_path, + b.sample_uid, + b.op_seq_bucket_id, + grp.group_uid +FROM graph_sample g +JOIN graph_net_sample_buckets b ON g.uuid = b.sample_uid +JOIN graph_net_sample_groups grp ON g.uuid = grp.sample_uid +WHERE g.deleted = 0 AND b.deleted = 0 AND grp.deleted = 0; + +select distinct group_concat(sample_uid) as sample_uids, op_seq_bucket_id +from ( + select * + from v_sample_buckets_groups + order by sample_uid asc +) +where sample_type = 'fusible_graph' +group by group_uid +order by op_seq_bucket_id asc +; diff --git a/graph_net/sqlite_util/select_representive_sample_uids.viba b/graph_net/sqlite_util/select_representive_sample_uids.viba new file mode 100644 index 000000000..cc433124e --- /dev/null +++ b/graph_net/sqlite_util/select_representive_sample_uids.viba @@ -0,0 +1,11 @@ +# select_representive_sample_uids.py +select_representive_sample_uids := + $ret list[$sample_uids str] + <- $total_sample_uids_and_op_seqs ArgParse[FileContentEachLine[$sample_uids str, TAB, $op_seq str]] + <- $selected_op_seqs ArgParse[FileContentEachLine[$op_seq str]] + # inline + <- ($grouped_by_op_seq dict[$op_seq str, list[$sample_uids str]] <- $total_sample_uids_and_op_seqs) + <- $get_max_len_sample_uids (str <- list[$sample_uids str]) + <- ($op_seq_to_max_len_sample_uids dict[$op_seq, $max_len_sample_uids str] + <- $get_max_len_sample_uids <- $grouped_by_op_seq) + <- ($ret <- $op_seq_to_max_len_sample_uids <- $selected_op_seqs)