happyz
/
llama.cpp
mirror of https://github.com/ggerganov/llama.cpp.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

llama-fit-params: fix step size for last device (#18415)

This commit is contained in:
Johannes Gäßler 2025-12-28 10:52:09 +01:00 committed by GitHub
parent e59efe6a78
commit f8d561eb87
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 3 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
src/llama.cpp
View File

@ -512,6 +512,9 @@ static void llama_params_fit_impl(
if (mem_high[id] > targets[id]) { if (mem_high[id] > targets[id]) {
assert(ngl_per_device_high[id].n_layer > ngl_per_device[id].n_layer); assert(ngl_per_device_high[id].n_layer > ngl_per_device[id].n_layer);
uint32_t delta = ngl_per_device_high[id].n_layer - ngl_per_device[id].n_layer; uint32_t delta = ngl_per_device_high[id].n_layer - ngl_per_device[id].n_layer;
if (hp_nex > 0 && size_t(id) == nd - 1) {
delta--;
}
LLAMA_LOG_DEBUG("%s: start filling device %" PRIu32 ", delta=%" PRIu32 "\n", __func__, id, delta); LLAMA_LOG_DEBUG("%s: start filling device %" PRIu32 ", delta=%" PRIu32 "\n", __func__, id, delta);
while (delta > 1) { while (delta > 1) {
uint32_t step_size = int64_t(delta) * (targets[id] - mem[id]) / (mem_high[id] - mem[id]); uint32_t step_size = int64_t(delta) * (targets[id] - mem[id]) / (mem_high[id] - mem[id]);