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

test zero max buffer size

Browse source
This commit is contained in:
slaren 2024-06-26 17:11:09 +02:00
parent 8854044561
commit 1e6e363d7f
2 changed files with 18 additions and 13 deletions
Download patch file Download diff file Expand all files Collapse all files

24
ggml-alloc.c
View file

@ -988,19 +988,19 @@ ggml_backend_buffer_t ggml_backend_alloc_ctx_tensors_from_buft(struct ggml_conte
this_size = GGML_PAD(ggml_backend_buft_get_alloc_size(buft, t), alignment);
}
if (this_size > max_size) {
fprintf(stderr, "%s: tensor %s is too large to fit in a %s buffer (tensor size: %zu, max buffer size: %zu)\n",
__func__, t->name,
ggml_backend_buft_name(buft),
this_size, max_size);
for (size_t i = 0; i < n_buffers; i++) {
ggml_backend_buffer_free(buffers[i]);
}
free(buffers);
return NULL;
}
//if (this_size > max_size) {
// fprintf(stderr, "%s: tensor %s is too large to fit in a %s buffer (tensor size: %zu, max buffer size: %zu)\n",
// __func__, t->name,
// ggml_backend_buft_name(buft),
// this_size, max_size);
// for (size_t i = 0; i < n_buffers; i++) {
// ggml_backend_buffer_free(buffers[i]);
// }
// free(buffers);
// return NULL;
//}
if ((cur_buf_size + this_size) > max_size) {
if (cur_buf_size != 0 && (cur_buf_size + this_size) > max_size) {
// allocate tensors in the current buffer
if (!alloc_tensor_range(ctx, first, t, buft, cur_buf_size, &buffers, &n_buffers)) {
return NULL;