sdgoij/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2026-05-13 04:24:17 +00:00
Code Issues Packages Projects Releases Wiki Activity

graph : make the compute graph constant with respect to active samplers

Browse Source
This commit is contained in:
Georgi Gerganov
2025-12-10 15:54:33 +02:00
parent 0ecee8be37
commit c02654eb7d
3 changed files with 48 additions and 36 deletions
Download Patch File Download Diff File Expand all files Collapse all files

15
src/llama-context.cpp
View File

@@ -1241,7 +1241,10 @@ static void copy_tensor_async_ints(
for (const auto & [seq_id, tensor] : tensor_map) {
auto it = seq_to_row.find(seq_id);
GGML_ASSERT(it != seq_to_row.end());
if (it == seq_to_row.end()) {
continue;
}
const uint32_t row = it->second;
GGML_ASSERT(row < sampled_size);
@@ -1265,7 +1268,10 @@ static void copy_tensor_async_floats(
for (const auto & [seq_id, tensor] : tensor_map) {
auto it = seq_to_row.find(seq_id);
GGML_ASSERT(it != seq_to_row.end());
if (it == seq_to_row.end()) {
continue;
}
const uint32_t row = it->second;
GGML_ASSERT(row < counts.size());
@@ -1293,7 +1299,10 @@ static void copy_tensor_async_candidates(
for (const auto & [seq_id, tensor] : tensor_map) {
auto it = seq_to_row.find(seq_id);
GGML_ASSERT(it != seq_to_row.end());
if (it == seq_to_row.end()) {
continue;
}
const uint32_t row = it->second;
GGML_ASSERT(row < counts.size());