vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Fix off by one error in dump_kv_cache_view

Browse source
This commit is contained in:
KerfuffleV2 2023-11-23 07:44:10 -07:00
parent 7688d7204f
commit bc1c346ae8
1 changed files with 8 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

9
common/common.cpp
View file

@ -1397,8 +1397,10 @@ void dump_kv_cache_view(const llama_kv_cache_view & view, int row_size) {
printf("=== Dumping KV cache. total cells %d, max sequences per cell %d, populated cells %d, total tokens in cache %d, largest empty slot=%d @ %d", printf("=== Dumping KV cache. total cells %d, max sequences per cell %d, populated cells %d, total tokens in cache %d, largest empty slot=%d @ %d",
view.n_cells, view.n_max_seq, view.used_cells, view.token_count, view.max_contiguous_cells, view.max_contiguous_cells_idx); view.n_cells, view.n_max_seq, view.used_cells, view.token_count, view.max_contiguous_cells, view.max_contiguous_cells_idx);
llama_kv_cache_view_cell * c_curr = view.cells; llama_kv_cache_view_cell * c_curr = view.cells;
struct llama_kv_cache_view_cell_sequence * cs_curr = view.cells_sequences; struct llama_kv_cache_view_cell_sequence * cs_curr = view.cells_sequences;
for (int i = 0; i < view.n_cells; i++, c_curr++, cs_curr += view.n_max_seq) { for (int i = 0; i < view.n_cells; i++, c_curr++, cs_curr += view.n_max_seq) {
if (i % row_size == 0) { if (i % row_size == 0) {
printf("\n%5d: ", i); printf("\n%5d: ", i);
@ -1407,19 +1409,22 @@ void dump_kv_cache_view(const llama_kv_cache_view & view, int row_size) {
for (int j = 0; j < view.n_max_seq; j++) { for (int j = 0; j < view.n_max_seq; j++) {
if (cs_curr[j].seq_id >= 0) { seq_count++; } if (cs_curr[j].seq_id >= 0) { seq_count++; }
} }
putchar(slot_chars[std::min(sizeof(slot_chars) - 1, size_t(seq_count))]); putchar(slot_chars[std::min(sizeof(slot_chars) - 2, size_t(seq_count))]);
} }
printf("\n=== Done dumping\n"); printf("\n=== Done dumping\n");
} }
void dump_kv_cache_view_seqs(const llama_kv_cache_view & view, int row_size) { void dump_kv_cache_view_seqs(const llama_kv_cache_view & view, int row_size) {
static const char slot_chars[] = "0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz"; static const char slot_chars[] = "0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz";
printf("=== Dumping KV cache. total cells %d, max sequences per cell %d, populated cells %d, total tokens in cache %d, largest empty slot=%d @ %d\n", printf("=== Dumping KV cache. total cells %d, max sequences per cell %d, populated cells %d, total tokens in cache %d, largest empty slot=%d @ %d\n",
view.n_cells, view.n_max_seq, view.used_cells, view.token_count, view.max_contiguous_cells, view.max_contiguous_cells_idx); view.n_cells, view.n_max_seq, view.used_cells, view.token_count, view.max_contiguous_cells, view.max_contiguous_cells_idx);
std::unordered_map<llama_seq_id, size_t> seqs; std::unordered_map<llama_seq_id, size_t> seqs;
llama_kv_cache_view_cell * c_curr = view.cells; llama_kv_cache_view_cell * c_curr = view.cells;
struct llama_kv_cache_view_cell_sequence * cs_curr = view.cells_sequences; struct llama_kv_cache_view_cell_sequence * cs_curr = view.cells_sequences;
for (int i = 0; i < view.n_cells; i++, c_curr++, cs_curr += view.n_max_seq) { for (int i = 0; i < view.n_cells; i++, c_curr++, cs_curr += view.n_max_seq) {
for (int j = 0; j < view.n_max_seq; j++) { for (int j = 0; j < view.n_max_seq; j++) {
if (cs_curr[j].seq_id < 0) { continue; } if (cs_curr[j].seq_id < 0) { continue; }
@ -1430,6 +1435,7 @@ void dump_kv_cache_view_seqs(const llama_kv_cache_view & view, int row_size) {
} }
if (seqs.size() + 1 >= sizeof(slot_chars)) { break; } if (seqs.size() + 1 >= sizeof(slot_chars)) { break; }
} }
printf("=== Sequence legend: "); printf("=== Sequence legend: ");
for (const auto & it : seqs) { for (const auto & it : seqs) {
printf("%zu=%d, ", it.second, it.first); printf("%zu=%d, ", it.second, it.first);
@ -1452,5 +1458,6 @@ void dump_kv_cache_view_seqs(const llama_kv_cache_view & view, int row_size) {
} }
putchar(' '); putchar(' ');
} }
printf("\n=== Done dumping\n"); printf("\n=== Done dumping\n");
} }